system

Abstract

A system is provided.SOLUTION: A system including means for collecting data from a database or an external API in order to acquire data to be collected, means for normalizing the collected data, unifying a format of the data, and complementing a missing value, means for detecting personal information from the normalized data, means for detecting a trade secret included in the data, means for evaluating whether or not there is a bias in the collected data, means for generating a report based on a result of the evaluation, and means for notifying a user of the generated report.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The technology of the present disclosure relates to a system. [Background technology]

[0002] Patent document 1 discloses a persona chatbot control method performed by at least one processor, the method including the steps of receiving a user utterance, adding the user utterance to a prompt including an instruction sentence related to a description of the chatbot character, encoding the prompt, and inputting the encoded prompt into a language model to generate a chatbot utterance in response to the user utterance. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-180282 Summary of the Invention [Problem to be solved by the invention]

[0004] As AI technology advances, many companies are increasingly using data for AI training. However, this data contains personal information and trade secrets, which poses the risk of information leaks and legal issues. Furthermore, bias in the training data can affect the performance and fairness of AI models. Therefore, a system is needed to evaluate and examine problems contained in AI training data in advance and reduce risks. [Means for solving the problem]

[0005] This invention provides a system for evaluating and scrutinizing AI training data. Specifically, the system includes: means for collecting data from databases and external APIs to acquire the target data; means for normalizing the collected data, standardizing the data format, and imputing missing values; means for detecting personal information from the normalized data; means for detecting trade secrets contained in the data; means for evaluating whether the collected data is biased; means for generating a report based on the results of the evaluation; and means for notifying users of the generated report. This enables companies to identify potential risks in AI training data in advance and take appropriate measures.

[0006] "Collection means" refers to the means for obtaining training data from a database or external API.

[0007] "Normalization methods" are methods for standardizing the format of collected data and filling in missing values.

[0008] The "personal information detection means" is a means for identifying personal information such as name, address, telephone number, and email address from normalized data.

[0009] The "trade secret detection means" is a means for identifying trade secrets such as development information for specific products and in-house project names contained in the data.

[0010] "Bias assessment methods" are methods for statistically assessing whether collected data is biased in terms of gender, age, region, etc.

[0011] The "report generation means" is a means for creating a report based on the evaluation results regarding personal information, trade secrets, and data bias.

[0012] The "notification means" is a means for notifying the user of the generated report. [Brief explanation of the drawings]

[0013] [Figure 1] 1 is a conceptual diagram showing an example of the configuration of a data processing system according to a first embodiment. [Figure 2] 1 is a conceptual diagram showing an example of main functions of a data processing device and a smart device according to a first embodiment. [Figure 3] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a second embodiment. [Figure 4] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and smart glasses according to a second embodiment. [Figure 5] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a third embodiment. [Figure 6] FIG. 11 is a conceptual diagram showing an example of main functions of a data processing device and a headset-type terminal according to a third embodiment. [Figure 7] FIG. 10 is a conceptual diagram showing an example of the configuration of a data processing system according to a fourth embodiment. [Figure 8] FIG. 10 is a conceptual diagram showing an example of main functions of a data processing device and a robot according to a fourth embodiment. [Figure 9] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 10] 1 shows an emotion map onto which multiple emotions are mapped. [Figure 11] FIG. 3 is a sequence diagram showing a processing flow of the data processing system according to the first embodiment. [Figure 12] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 1. [Figure 13] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system according to the second embodiment when an emotion engine is combined. [Figure 14] FIG. 10 is a sequence diagram showing the flow of processing in the data processing system in Application Example 2 when an emotion engine is combined. DETAILED DESCRIPTION OF THE INVENTION

[0014] An example of an embodiment of a system according to the technology of the present disclosure will be described below with reference to the accompanying drawings.

[0015] First, the terms used in the following description will be explained.

[0016] In the following embodiments, a coded processor (hereinafter simply referred to as a "processor") may be a single arithmetic device or a combination of multiple arithmetic devices. Furthermore, a processor may be a single type of arithmetic device or a combination of multiple types of arithmetic devices. Examples of arithmetic devices include a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a GPGPU (General-Purpose computing on Graphics Processing Units), and an APU (Accelerated Processing Unit).

[0017] In the following embodiments, a coded RAM (Random Access Memory) is a memory in which information is temporarily stored and is used as a working memory by a processor.

[0018] In the following embodiments, the coded storage is one or more non-volatile storage devices that store various programs, various parameters, etc. Examples of non-volatile storage devices include flash memory (SSD (Solid State Drive)), magnetic disks (e.g., hard disks), and magnetic tapes.

[0019] In the following embodiments, a communication I / F (Interface) with a symbol is an interface including a communication processor, an antenna, etc. The communication I / F controls communication between multiple computers. Examples of communication standards applied to the communication I / F include wireless communication standards including 5G (5th Generation Mobile Communication System), Wi-Fi (registered trademark), Bluetooth (registered trademark), etc.

[0020] In the following embodiments, "A and / or B" is synonymous with "at least one of A and B." In other words, "A and / or B" means that it may be only A, only B, or a combination of A and B. Furthermore, in this specification, the same concept as "A and / or B" is also applied when three or more things are expressed connected by "and / or."

[0021] [First embodiment]

[0022] FIG. 1 shows an example of the configuration of a data processing system 10 according to the first embodiment.

[0023] 1, a data processing system 10 includes a data processing device 12 and a smart device 14. An example of the data processing device 12 is a server.

[0024] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0025] The smart device 14 includes a computer 36, a reception device 38, an output device 40, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The reception device 38, the output device 40, and the camera 42 are also connected to the bus 52.

[0026] The reception device 38 includes a touch panel 38A, a microphone 38B, and the like, and receives user input. The touch panel 38A detects contact with an indicator (for example, a pen or a finger) to receive user input by the touch of the indicator. The microphone 38B detects the user's voice to receive user input by voice. The control unit 46A transmits data indicating the user input received by the touch panel 38A and the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the data indicating the user input.

[0027] The output device 40 includes a display 40A and a speaker 40B, and presents data to the user 20 by outputting the data in a form of expression that the user 20 can perceive (for example, audio and / or text). The display 40A displays visible information such as text and images in accordance with instructions from the processor 46. The speaker 40B outputs audio in accordance with instructions from the processor 46. The camera 42 is a compact digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor.

[0028] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 control the exchange of various information between the processor 46 and the processor 28 via the network 54.

[0029] FIG. 2 shows an example of the main functions of the data processing device 12 and the smart device 14.

[0030] 2, in the data processing device 12, a specific process is performed by the processor 28. A specific processing program 56 is stored in the storage 32. The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific process is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0031] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0032] In the smart device 14, the processor 46 performs the reception output process. The storage 50 stores a reception output program 60. The reception output program 60 is used in conjunction with the specific processing program 56 by the data processing system 10. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0033] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0034] This invention relates to a system for evaluating and examining AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[0035] Overall Overview

[0036] The server collects the necessary training data from the company's internal database or external API, normalizes it, detects personal information and trade secrets from the normalized data, and evaluates the data for bias. A report is generated based on the evaluation results, and the server notifies the user of the report.

[0037] Details of each step

[0038] Data collection

[0039] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[0040] Data normalization

[0041] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[0042] Personal information detection

[0043] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[0044] Trade Secret Detection

[0045] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[0046] Assessing data bias

[0047] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[0048] Generate reports

[0049] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias. This report includes the detected issues and how to address them. For example, the report is generated in PDF or HTML format and the contents are summarized.

[0050] User Notification

[0051] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[0052] Specific examples

[0053] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. A report is generated based on the evaluation results, and the report is notified to a person in charge within the company. The person in charge reviews the report and takes necessary measures, enabling the development of a reliable AI model.

[0054] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[0055] The processing flow will be explained below.

[0056] Step 1: Collect data

[0057] The server collects learning data from the company's internal databases and external APIs.

[0058] Action 1.1: Obtain database connection information and establish a connection.

[0059] Action 1.2: Execute a query on the data to be collected and retrieve the data. For example, query customer purchase history data.

[0060] Action 1.3: Save the acquired data to temporary storage.

[0061] Step 2: Normalize the data

[0062] The server normalizes the collected data.

[0063] Action 2.1: Standardize data formats, such as date formats and monetary values.

[0064] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[0065] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[0066] Step 3: Discover personal information

[0067] The server detects personal information from the normalized data.

[0068] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[0069] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[0070] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[0071] Step 4: Discover trade secrets

[0072] The server detects trade secrets contained in the data.

[0073] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[0074] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[0075] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[0076] Step 5: Assess data for bias

[0077] The server evaluates the collected data for bias.

[0078] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[0079] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[0080] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[0081] Step 6: Generate reports

[0082] The server generates a report based on the above evaluation results.

[0083] Action 6.1: Compile a list of discovered personal information and trade secrets.

[0084] Action 6.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[0085] Action 6.3: Write a report detailing the problem and its solution.

[0086] Action 6.4: Output the completed report in PDF or HTML format and save it on the server.

[0087] Step 7: Notify users

[0088] The server notifies the user of the generated report.

[0089] Action 7.1: Create a notification to let the user know where to save the report.

[0090] Action 7.2: Send a notification to the user via email or corporate notification system.

[0091] Action 7.3: Provide an interface for receiving feedback from the user.

[0092] Example 1

[0093] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0094] Conventional data collection and evaluation systems often failed to properly normalize the collected data, and it was difficult to detect personal or confidential information. Furthermore, evaluating whether the data was biased and generating reports based on the evaluation results was time-consuming and prone to human error. Furthermore, notifications of generated reports to users were often done manually, which was inefficient. A highly reliable system was needed to solve these problems.

[0095] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0096] In this invention, the server includes means for collecting information from a database or external API to obtain information to be collected, means for normalizing the collected information, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized information, means for detecting confidential information contained in the information, means for evaluating whether the collected information is biased, means for generating a report based on the results of the above evaluations, and means for notifying the user of the generated report.This enables efficient data collection, evaluation, reporting, and notification while increasing the accuracy of the data.

[0097] A "data collection target" is a set of information required for analysis or evaluation.

[0098] A "database" is a digital structure that organizes information and allows it to be searched and updated.

[0099] An "external API" is an interface for accessing external systems and services to obtain data.

[0100] "Information" means data or textual content that is collected, processed, and evaluated for a specific purpose or use.

[0101] "Normalization" is the process of standardizing data formats, removing unnecessary data, and filling in missing values.

[0102] A "missing value" is a value that is not recorded or is missing in a dataset.

[0103] "Personal information" means information that can be used to identify a specific individual, such as name, address, or telephone number.

[0104] "Confidential Information" is non-public, sensitive data that is specific to a particular organization or project.

[0105] "Bias" is a state in which data is overly concentrated or absent in a particular direction or attribute.

[0106] A "report" is a documented result of a particular analysis or evaluation, in a format that allows a user to review the contents.

[0107] "User" is a person or organization that receives reports generated by this system and takes action or makes decisions based on them.

[0108] This invention relates to a system for evaluating and examining AI learning data, and is implemented primarily using a server. Below, the program processing of this system is explained in natural language.

[0109] First, the server uses a database or external API to obtain the information to be collected. Specifically, data is collected using an SQL database (e.g., MySQL, PostgreSQL) or a NoSQL database (e.g., MongoDB). For example, a company's server collects customer purchase history data and temporarily stores it in storage (e.g., Amazon S3).

[0110] The server then normalizes the collected data. This process uses data processing libraries such as Apache Spark and Pandas. Data quality is improved by standardizing data formats, imputing missing values, and removing duplicate data. For example, standardizing date formats and replacing NULL values ​​with a mean value are performed.

[0111] Next, the server detects personal information from the normalized data using regular expressions (e.g., Python's re module) or machine learning models (e.g., spaCy, NLTK). For example, personal information such as name, address, phone number, and email address is detected and temporarily stored as a list.

[0112] Furthermore, the server detects confidential information contained in the data. It extracts confidential information using specific keyword lists and machine learning models (e.g., BERT, GPT). For example, project names and product development information are identified and stored as a list.

[0113] The server then assesses the collected data for bias using Matplotlib and statistical methods (e.g., t-tests, chi-squared tests), visually assessing data distribution with histograms and scatter plots, and performing statistical tests to determine bias.

[0114] Based on the results of the evaluation, the server generates a report in PDF (e.g., ReportLab) or HTML (e.g., Jinja2 template engine) format that includes the issues found and how to address them.

[0115] Finally, the server notifies the user of the generated report via email (e.g., SMTP server) or an internal corporate notification system (e.g., Slack, Microsoft Teams). For example, a link indicating where the generated report is saved may be sent via email.

[0116] Specific examples

[0117] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects purchase history data from a database (e.g., MySQL). Next, it normalizes the collected data using Pandas and detects personal or confidential information using regular expressions and spaCy. It then visualizes the data distribution using Matplotlib and performs statistical tests to evaluate data bias. Based on the evaluation results, it generates a report using ReportLab, and the report is sent to the responsible person via email via an SMTP server. The responsible person reviews the report and takes any necessary measures, enabling the development of a reliable AI model.

[0118] Prompt Sentence Examples

[0119] "Create a program to collect customer purchase history data, normalize the data, and detect personal and confidential information. Also, evaluate the data for bias, generate a PDF report, and notify the user of the report via email."

[0120] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[0121] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0122] Step 1:

[0123] The server collects information from a database or external API. It takes a query or API request as input and the collected data as output. For example, to get a user's purchasing history, it runs an SQL query to retrieve the information from the database. The server then temporarily stores this data in storage.

[0124] Step 2:

[0125] The server normalizes the collected data. It requires the collected raw data as input and obtains normalized data as output. Specific operations include standardizing data formats, imputing missing values, and removing duplicate data using Pandas. For example, it standardizes date formats and imputes NULL values ​​with the average value.

[0126] Step 3:

[0127] The server detects personal information from the normalized data. Normalized data is required as input, and the detected personal information is listed as output. Specifically, it uses regular expressions and spaCy to identify personal information such as name, address, phone number, and email address. This information is temporarily stored in storage.

[0128] Step 4:

[0129] The server detects confidential information contained in the data. Normalized data is required as input, and the detected confidential information is listed as output. Specifically, it uses BERT and a specific keyword list to extract confidential information such as project names and product development information. This information is also temporarily stored in storage.

[0130] Step 5:

[0131] The server evaluates whether the collected data is biased. Normalized data is required as input, and the output is the bias evaluation result. Specifically, Matplotlib is used to visually evaluate the data distribution using histograms and scatter plots, and statistical tests such as the chi-square test are performed. The evaluation results are compiled into a report.

[0132] Step 6:

[0133] The server generates a report based on the results of the assessment of detected personal information, confidential information, and bias. It takes as input the results of each detection and assessment, and as output the generated report. It uses ReportLab or the Jinja2 template engine to create reports in PDF and HTML format.

[0134] Step 7:

[0135] The server notifies the user of the generated report. The generated report and the user's contact information are required as input, and a notification is sent to the user as output. Specifically, it sends an email via an SMTP server and provides a link to the location where the report is saved. An in-house notification system can also be used as a notification method.

[0136] (Application example 1)

[0137] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0138] Sensor data and camera video data collected by autonomous vehicles may contain personal information and trade secrets, so systems are needed to properly process this data while ensuring data accuracy and compliance. Furthermore, if the collected data is biased, it can affect the learning results of the AI ​​model, reducing safety and reliability. Especially for autonomous vehicles, data bias can have a direct impact on the vehicle's driving behavior, so it is necessary to evaluate whether or not there is data bias and correct it appropriately.

[0139] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0140] In this invention, the server includes means for collecting data from a database or an external API to acquire data to be collected, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for collecting and normalizing sensor data and camera video data from an autonomous vehicle and evaluating whether there is bias in the personal information and data, means for generating a report based on the above evaluation results, and means for notifying a user of the generated report. This makes it possible to evaluate data bias and develop and operate a highly reliable AI model while ensuring data accuracy and compliance in autonomous vehicles.

[0141] A "database" is a set of structured data that allows for efficient storage, management, retrieval, and updating of information.

[0142] An "external API" is an interface for communicating with and exchanging data with other systems or services.

[0143] "Normalization" is the process of standardizing the format of data and imputing missing values.

[0144] "Personal information" is information that can identify a specific individual (e.g., name, address, telephone number, email address).

[0145] "Trade secrets" are information that is kept secret in order to maintain a company's competitive edge, and include development information for specific products and internal project names.

[0146] "Data bias" refers to a situation in which collected data is unevenly distributed across certain attributes (e.g., gender, age, region).

[0147] "Sensor data" refers to data collected from various sensors (e.g., GPS, accelerometer, camera, etc.).

[0148] "Camera video data" refers to still images and video data captured by a camera.

[0149] A "report" is a document that summarizes the results of a particular evaluation or analysis.

[0150] "User" means a person or entity that uses the System.

[0151] An "on-board computer" is a computing device installed in a vehicle and used to process data from sensors and cameras.

[0152] To implement the present invention, a system can be constructed in the following manner: The server performs processes for acquiring, normalizing, detecting, evaluating, generating reports, and notifying data to be collected.

[0153] Program processing and use of hardware and software

[0154] 1. Data Collection

[0155] The server uses a database and external API to collect sensor data and camera video data from the autonomous vehicle, allowing it to understand the vehicle's operating status in real time. For example, it uses the requests library to send HTTP requests and retrieve data.

[0156] 2. Data normalization

[0157] The server performs data normalization to standardize the collected data into a consistent format and impute missing values, using libraries such as pandas and sklearn.preprocessing, including standardizing date formats and scaling sensor data.

[0158] 3. Personal information detection

[0159] The server uses regular expressions and machine learning models to detect personal information from the normalized data, specifically using the re library to identify names, addresses, phone numbers, email addresses, etc.

[0160] 4. Trade Secret Discovery

[0161] The data collected by companies may contain trade secrets, which the server detects using a specific list of keywords and machine learning models.

[0162] 5. Assessing data bias

[0163] The server aggregates statistical information such as gender, age, and region to assess whether the collected data is biased, and performs statistical tests. The pandas library is used to generate histograms and scatter plots to visually evaluate the data distribution.

[0164] 6. Report Generation

[0165] Based on the results of the assessment of detected personal information, trade secrets, and data bias, the server generates a report in PDF or HTML format detailing the issues and how to address them.

[0166] 7. Notice to Users

[0167] Once a report is generated, the user is notified via email or the company's internal notification system, for example by sending a link to the location where the report is saved.

[0168] Specific examples

[0169] Below are some specific examples of data collection and analysis in autonomous vehicles.

[0170] 1. Example prompt:

[0171] The API collects sensor data and camera video data from an autonomous vehicle, retrieves the data every 10 seconds, and stores it on the vehicle's internal computer. The API endpoints are provided below (sensor: http: / / example.com / sensor_data, camera: http: / / example.com / camera_data).

[0172]

[0173] Scale the collected data and standardize the date format to 'YYYY-MM-DD HH:MM:SS'.

[0174]

[0175] Use regular expressions to detect and list names, phone numbers, and email addresses in your data.

[0176]

[0177] Use gender and age demographics to assess whether your dataset is biased.

[0178] In this way, the system can support the development and operation of reliable AI models while ensuring a high level of security and compliance for data collected from autonomous vehicles.

[0179] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0180] Step 1:

[0181] The server collects sensor data and camera video data from autonomous vehicles from databases and external APIs.

[0182] Input: Database or API endpoint.

[0183] Data processing: Data is retrieved using HTTP requests and temporarily stored in the server's storage. Specifically, data is retrieved in real time using the requests library.

[0184] Output: The raw data collected.

[0185] Step 2:

[0186] The server normalizes the collected data, standardizes it into a consistent format, and imputes missing values.

[0187] Input: Raw data collected.

[0188] Data processing: Using pandas and the sklearn.preprocessing library, we standardized date formats, scaled the data, and imputed missing values ​​with the mean and median.

[0189] Specifically, a data frame is created and the data format of each column is aligned.

[0190] Output: Normalized data.

[0191] Step 3:

[0192] The server detects personal information from the normalized data.

[0193] Input: Normalized data.

[0194] Data processing: Use the re library to extract names, addresses, phone numbers, and email addresses using regular expressions.

[0195] Specifically, pattern matching is performed to create a list of personal information.

[0196] Output: A list of personal information found.

[0197] Step 4:

[0198] The server detects trade secrets from the normalized data.

[0199] Input: Normalized data.

[0200] Data processing: Detecting specific information using company-defined keyword lists and machine learning models.

[0201] Specifically, it loads a machine learning model and makes predictions on the data.

[0202] Output: A list of discovered trade secrets.

[0203] Step 5:

[0204] The server evaluates the collected data for bias.

[0205] Input: Normalized data.

[0206] Data processing: Use the pandas library to aggregate statistics and evaluate the distribution of attribute values ​​such as gender and age.

[0207] Specifically, histograms and scatter plots are generated and biases are detected using statistical methods.

[0208] Output: Report of bias assessment.

[0209] Step 6:

[0210] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias.

[0211] Input: List of personal information, list of trade secrets, bias assessment report.

[0212] Data processing: Compile this information into a document in PDF or HTML format.

[0213] Specifically, data is inserted into an automatically generated template to generate a document.

[0214] Output: Compliance report.

[0215] Step 7:

[0216] The server notifies the user of the generated report.

[0217] Input: Compliance report.

[0218] Data processing: Send a link to the report to users via email or internal notification systems.

[0219] Specifically, it uses an SMTP server to send an email and includes a link.

[0220] Output: Notification email to user.

[0221] Furthermore, an emotion engine that estimates the user's emotion may be combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59 and perform identification processing using the user's emotion.

[0222] This invention combines an emotion engine with a system that evaluates and examines AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[0223] Overall Overview

[0224] The server collects the necessary learning data from internal company databases and external APIs and normalizes it. It then detects personal information and trade secrets from the normalized data and evaluates the data for bias. It generates a report based on the evaluation results and notifies the user of the report. The server also incorporates an emotion engine that recognizes user emotions based on the collected data and analyzes user feedback and responses.

[0225] Details of each step

[0226] Data collection

[0227] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[0228] Data normalization

[0229] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[0230] Personal information detection

[0231] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[0232] Trade Secret Detection

[0233] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[0234] Assessing data bias

[0235] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[0236] Emotion recognition by emotion engine

[0237] The server recognizes the user's emotions based on the collected data. This emotion recognition involves analyzing voice data and text data. For example, the server analyzes voice data to identify the user's emotions (happiness, sadness, anger, etc.). It also analyzes text data using natural language processing (NLP) techniques to identify the user's emotions.

[0238] Generate reports

[0239] The server generates a report based on the results of the user sentiment analysis, as well as the results of the evaluation of detected personal information and trade secrets, and data bias. This report includes detected problems and how to address them, as well as user feedback based on the sentiment analysis. For example, the report is generated in PDF or HTML format and the contents are summarized.

[0240] User Notification

[0241] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[0242] Specific examples

[0243] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the content. Finally, this report is notified to a person in the company who reviews it and takes necessary measures, enabling the development of a reliable AI model.

[0244] This system allows companies to identify potential risks in AI learning data in advance and take appropriate measures, and by taking user emotions into consideration, it is possible to obtain more accurate analysis results. This makes it possible to develop highly reliable AI and significantly reduce operational risks.

[0245] The processing flow will be explained below.

[0246] Step 1: Collect data

[0247] The server collects learning data from the company's internal databases and external APIs.

[0248] Action 1.1: Obtain database connection information and establish a connection.

[0249] Action 1.2: Query the data to be collected and retrieve the data. For example, query customer purchase history data.

[0250] Action 1.3: Save the acquired data to temporary storage.

[0251] Step 2: Normalize the data

[0252] The server normalizes the collected data.

[0253] Action 2.1: Standardize data formats, such as date formats and monetary values.

[0254] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[0255] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[0256] Step 3: Discover personal information

[0257] The server detects personal information from the normalized data.

[0258] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[0259] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[0260] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[0261] Step 4: Discover trade secrets

[0262] The server detects trade secrets contained in the data.

[0263] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[0264] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[0265] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[0266] Step 5: Assess data for bias

[0267] The server evaluates the collected data for bias.

[0268] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[0269] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[0270] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[0271] Step 6: Emotion Recognition with the Emotion Engine

[0272] The server recognizes the user's emotions based on the collected data.

[0273] Action 6.1: Analyze audio data. For example, take an audio file and identify emotions such as joy, sadness, or anger from the audio.

[0274] Activity 6.2: Analyze text data, for example customer reviews and feedback, using natural language processing techniques to identify sentiment.

[0275] Action 6.3: List the sentiment analysis results and save them in temporary storage.

[0276] Step 7: Generate reports

[0277] The server generates a report based on the above evaluation results.

[0278] Action 7.1: Compile a list of discovered personal information and trade secrets.

[0279] Action 7.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[0280] Action 7.3: Summarize user feedback based on sentiment analysis results.

[0281] Action 7.4: Output the completed report in PDF or HTML format and save it on the server.

[0282] Step 8: Notify users

[0283] The server notifies the user of the generated report.

[0284] Action 8.1: Create a notification to let the user know where the report is saved.

[0285] Action 8.2: Send a notification to the user via email or corporate notification system.

[0286] Action 8.3: Provide an interface for receiving feedback from the user.

[0287] Example 2

[0288] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0289] Conventional data collection and analysis systems have had problems such as difficulty in identifying personal information or trade secrets contained in the collected training data, and insufficient evaluation of data bias. Furthermore, because data is evaluated without taking user emotions into consideration, the accuracy of the final analysis results is likely to be reduced.

[0290] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0291] In this invention, the server includes means for collecting data from databases and external information sources, means for normalizing the collected data, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for analyzing voice data and text data to recognize user emotions, means for generating a report based on the above evaluation results and emotion recognition results, and means for notifying the user of the generated report. This makes it possible to identify potential risks in the data in advance, take appropriate measures, and perform highly accurate analysis that takes user emotions into consideration.

[0292] A "database" is a computer system that systematically stores information and allows it to be efficiently searched, manipulated, and updated.

[0293] An "external source" is a data source that resides outside the enterprise and is typically accessed via an API.

[0294] "Normalization" is the process of standardizing the format of collected data and filling in missing values.

[0295] "Personal information" is information that can identify a specific individual, and includes, for example, name, address, telephone number, and email address.

[0296] "Trade secrets" are important, non-public information about a company's business, including, for example, information about new products under development and details of internal projects.

[0297] "Data bias" refers to imbalances in a dataset based on specific attributes or conditions.

[0298] "Emotion recognition" is the process of analyzing voice and text data to identify a user's emotions (e.g., joy, sadness, anger, etc.).

[0299] A "report" is a document summarizing detected personal information, trade secrets, data bias, and emotion recognition results.

[0300] "Notification" refers to the act of informing the user of the contents of the generated report, and is carried out via email or an in-company notification system.

[0301] This invention relates to a system that automatically performs processes from data collection to analysis and report generation. This system will be described in detail below.

[0302] Data collection

[0303] The server has a means for collecting training data from the company's internal database and external information sources. For example, the server executes SQL queries against the database to obtain purchase history data. It also obtains data from external information sources using HTTP requests and performs authentication using an API key.

[0304] For example, to retrieve purchase history from the database, the server executes the following SQL query:

[0305] SELECT FROM PurchaseHistory WHERE Date >= '2023-01-01'

[0306] To collect data from an external API, send an HTTP request like this:

[0307] GET https: / / api.example.com / reviews?product_id=123

[0308] Data normalization

[0309] The server normalizes the collected data by standardizing the data format and imputing missing values, such as standardizing date formats to "YYYY-MM-DD," standardizing numeric display formats, and replacing NULL values ​​with the mean or median.

[0310] For example, to standardize date formats, do the following:

[0311] Convert 2023 / 01 / 01 to 2023-01-01

[0312] To fill in missing values, for example, if "purchase amount" is NULL, the overall average purchase amount is substituted.

[0313] Personal information and trade secret detection

[0314] The server uses regular expressions and machine learning models to detect personal information in the normalized data. For example, the regular expression to identify email addresses:

[0315] [a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,4}

[0316] Trade secret detection uses specific keyword lists and machine learning models, for example, scanning data for keywords such as "new product" or "project X."

[0317] Assessing data bias

[0318] The server evaluates the collected data for bias. This evaluation process involves aggregating demographic information such as age, gender, and region, and performing statistical tests. For example, chi-square tests and t-tests are used to verify statistically significant bias. This process can use Python's Matplotlib library to generate histograms and scatter plots.

[0319] emotion recognition

[0320] The server analyzes the voice and text data to recognize the user's emotions. This emotion recognition uses a speech recognition engine (such as the Google Cloud Speech-to-Text API) to convert voice data into text, and natural language processing (NLP) technology to analyze text data.

[0321] As a concrete example, we use Waveform Transform and Google Cloud Speech-to-Text API to analyze audio files and identify user emotions.

[0322] Report generation and notification

[0323] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. This report is generated in PDF or HTML format and saved on the server. For example, the Python ReportLab library is used to generate the PDF report.

[0324] The generated report will be sent to the user via email or the company's internal notification system, for example, a link to the location where the generated report is saved will be sent via email:

[0325] http: / / example.com / reports / Report2023-10-01.pdf

[0326] Specific examples

[0327] For example, suppose a company wants to develop a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Then, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the results. Finally, the report is sent to a person in the company who reviews it and takes any necessary measures.

[0328] Prompt Sentence Examples

[0329] For example, here is an example of a prompt sentence being input to a generative AI model:

[0330] I want to create an AI model that uses customer purchase history data to infer user preferences and interests and propose personalized marketing strategies. How can I build a system that collects data, normalizes it, detects personal information and trade secrets, evaluates data bias, recognizes emotions using an emotion engine, generates reports, and notifies users?

[0331] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0332] Step 1: Collect data

[0333] The server collects data from a company's internal database or external information source. The server first connects to the database and uses authentication information to confirm access rights. For example, it executes an SQL query to retrieve purchase history data. The input is the database connection information and SQL query, and the output is a record of purchase history data. When collecting data from an external information source, the server authenticates using an API key and sends an API request. Specifically, the server sends an HTTP request and temporarily stores the data obtained as a response in storage on the server.

[0334] Step 2: Normalize the data

[0335] The server normalizes the collected data. In this process, the data format is standardized and missing values ​​are filled. For example, the date format is standardized to "YYYY-MM-DD" and NULL values ​​are replaced with the average or median. The input is the collected purchase history data, and the output is the data converted into a standardized format. Specifically, the server applies the rules for data format conversion and missing value filling, and temporarily stores the processed data.

[0336] Step 3: Discover personal information

[0337] The server detects personal information from the normalized data. Regular expressions and machine learning models are used for detection. The input is the normalized data, and the output is a list of detected personal information. Specifically, the server uses predefined regular expressions to identify email addresses and phone numbers, and uses machine learning models to identify addresses and names. The detection results are compiled into a list and stored in temporary storage.

[0338] Step 4: Discover trade secrets

[0339] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for a specific product or the names of internal projects. The input is normalized data, and the output is a list of detected trade secrets. Specifically, the server applies a keyword matching algorithm to extract relevant information. The detection results are listed and temporarily stored.

[0340] Step 5: Assess data for bias

[0341] The server evaluates whether the collected data is biased. This evaluation process involves compiling statistical information such as age, gender, and region, and performing statistical tests. The input is normalized data, and the output is a statistical report indicating whether bias is present or absent. Specifically, the server uses Python libraries to generate histograms and scatter plots and perform chi-square tests and t-tests. This allows the distribution of data to be visualized and statistical bias to be evaluated.

[0342] Step 6: Emotion Recognition

[0343] The server analyzes the voice and text data to recognize the user's emotions. Emotion recognition uses a voice recognition engine and natural language processing (NLP) technology. The input is voice and text data, and the output is analyzed emotion data. Specifically, the server converts the voice data into text (using the Google Cloud Speech-to-Text API) and analyzes the text using a sentiment analysis model. The analysis results are stored in a database.

[0344] Step 7: Generate reports

[0345] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. The input is all of these detection and assessment results, and the output is a report in PDF or HTML format. Specifically, the server uses the Python ReportLab library to generate a PDF report and summarize the report contents. The generated report is saved on the server.

[0346] Step 8: Notify users

[0347] The server notifies the user of the generated report. The input is the generated report and the user's contact information, and the output is a confirmation that the user has been notified. Specifically, the server retrieves the user's email address from a database, generates an email, and sends it using an SMTP server. The email includes a link indicating where the report is saved.

[0348] (Application example 2)

[0349] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart device 14 will be referred to as a "terminal."

[0350] In autonomous vehicles, in addition to analyzing driving data and evaluating bias, it is necessary to properly analyze the emotional feedback of the user while driving in order to optimize safety and operation management. Conventional systems lack the functionality to properly evaluate and process this emotional data, which has led to problems such as being unable to adequately evaluate bias in driving data and user stress.

[0351] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0352] In this invention, the server includes means for collecting data from databases and external APIs, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for incorporating an emotion engine that recognizes user emotions, means for generating a report based on the evaluation results and emotion recognition results, and means for notifying the user of the generated report. This enables comprehensive analysis of autonomous vehicle driving data and user emotional feedback, enabling optimization of safety and operation management.

[0353] A "database" is an organized collection of data, a system that stores various data and allows it to be easily searched, retrieved, and analyzed.

[0354] An "external API" is a program interface provided for linking with other software or systems, and is a means of obtaining and manipulating data.

[0355] "Normalization" is the process of standardizing the format of collected data and imputing missing values, improving data quality and consistency.

[0356] "Personal information" is information that identifies a specific individual, such as name, address, and telephone number.

[0357] A "trade secret" is non-public information held by a company or organization, such as technical or commercial information that is kept secret in order to maintain a competitive advantage.

[0358] Bias assessment is the process of detecting and analyzing biases and imbalances present in collected data.

[0359] An "emotion engine" is a technology for recognizing and analyzing a user's emotions, and is a system that estimates their emotional state from voice and text data.

[0360] A "report" is a report summarizing the results of analysis and evaluation of collected data, and is provided to users.

[0361] "Notification" is the process of communicating generated reports and important information to users, typically via email or smartphone notification systems.

[0362] This invention relates to a system for comprehensively analyzing driving data of autonomous vehicles and emotional feedback from users to optimize safety and operation management. The system includes a server, data collection means, data normalization means, personal information detection means, trade secret detection means, bias evaluation means, an emotion recognition engine, report generation means, and user notification means.

[0363] Server Configuration

[0364] The server collects various sensor data (e.g., GPS, speed, battery status, etc.) sent from the autonomous vehicle via a database or external API. The collected data is temporarily stored on the server. Data collection methods include scripts using programming languages ​​such as Python and libraries that call external APIs. For example, the requests library can be used to obtain the required data from external APIs.

[0365] Data normalization measures

[0366] Data normalization involves standardizing the format of the collected data and imputing missing values. To do this, we use the pandas library to manipulate data frames and perform the necessary format conversions and imputation of missing values. For example, we standardize date formats and replace NULL values ​​with the average value.

[0367] Personal information detection method

[0368] Personal information detection techniques identify personal information from normalized data using regular expressions and machine learning models. For example, email addresses and phone numbers can be extracted using regular expressions, and names and addresses can be detected using machine learning models (e.g., models using the scikit-learn library).

[0369] Trade secret detection methods

[0370] The trade secret detection method detects trade secrets in data. It uses a keyword list specified by the company and a machine learning model to detect internal project names and technical information. This also uses the pandas library and natural language processing techniques (e.g., the nltk library and spaCy library).

[0371] Bias assessment measures

[0372] The bias assessment tool aggregates statistical information from the collected data to detect abnormal bias. For example, it creates histograms and scatter plots using the matplotlib and seaborn libraries and performs statistical tests.

[0373] Emotion Recognition Engine

[0374] The emotion recognition engine analyzes voice data and text feedback while driving to identify the user's emotions. To analyze voice data, it uses a voice recognition library (e.g., the SpeechRecognition library), and to analyze the emotions of text data, it uses the TextBlob or VADER library.

[0375] Report Generation Method

[0376] The report generation tool generates reports based on detected personal information, trade secrets, data bias, and sentiment analysis results. These reports are generated in PDF and HTML formats, and are visually easy to understand by utilizing the matplotlib and pdfkit libraries.

[0377] User notification method

[0378] The user notification means notifies the user of the generated report via email or in-app notification. Email is sent using the smtplib library, and a push notification library (e.g., Pushy) is used for in-app notifications.

[0379] Specific examples

[0380] For example, if an autonomous vehicle senses stress due to traffic conditions and finds bias in its driving data, the system will detect the bias and stress in the driving data and provide a detailed report and improvement measures to the operations manager.

[0381] Prompt Sentence Examples

[0382] Please explain in natural language the processes of the "Self-Driving Vehicle Emotion Feedback & Data Analysis System": [Self-Driving Vehicle Sensor Data], [Driving Data Normalization], [Personal Information Anonymization], [Data Bias Assessment], [Emotion Recognition], [Report Generation], and [User Notification].

[0383] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0384] Step 1:

[0385] The server collects sensor data from autonomous vehicles. Input data includes GPS data, speed data, battery data, etc., which are obtained from databases and external APIs. Specifically, the server makes API calls, retrieves data using the requests library, and temporarily stores it. The output is the collected raw data.

[0386] Step 2:

[0387] The server normalizes the data collected in step 1. The input data is raw data, which is stored in a data frame using the PANDAS library. Specific operations include standardizing date formats, converting NULL values ​​to averages, and removing duplicate data. The output is normalized data.

[0388] Step 3:

[0389] The server detects personal information from the normalized data. The input data is normalized data, and regular expressions and machine learning models from the scikit-learn library are used. Specifically, it identifies names, addresses, and phone numbers, creates a list, and saves it in temporary storage. The output is a list of detected personal information.

[0390] Step 4:

[0391] The server detects trade secrets contained in the data. The input data is normalized data, and uses a keyword list specified by the company and natural language processing technology (nltk library and spaCy library). Specifically, it detects specific products and project names, creates a list, and saves it in temporary storage. The output is a list of detected trade secrets.

[0392] Step 5:

[0393] The server evaluates the bias of the collected data. The input data is normalized data, and it is visualized using the matplotlib and seaborn libraries and subjected to statistical testing. Specific operations include creating histograms and scatter plots and determining statistical bias. The output is the bias evaluation results.

[0394] Step 6:

[0395] The server incorporates an emotion engine that recognizes the user's emotions. The input data is voice data and text data from the driver's seat, and it uses a voice recognition library, TextBlob, and VADER library. Specific operations include analyzing voice data, identifying emotions, and identifying emotions from text data. The output is the emotion analysis results.

[0396] Step 7:

[0397] The server generates a report based on the detected personal information, trade secrets, data bias evaluation results, and sentiment analysis results. The input data are the evaluation results obtained at each step, and it uses the matplotlib and pdfkit libraries. Specific operations include visualizing the evaluation results and generating reports in PDF and HTML format. The output is the generated report.

[0398] Step 8:

[0399] The server notifies the user of the generated report. The input data is the generated report, and the smtplib library is used to send emails or push notifications. Specific operations include creating a notification message and sending an email or in-app notification. The output indicates that the notification has been completed.

[0400] The specific processing unit 290 transmits the result of the specific processing to the smart device 14. In the smart device 14, the control unit 46A causes the output device 40 to output the result of the specific processing. The microphone 38B acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 38B to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0401] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0402] In the above embodiment, an example in which the specific process is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific process may be performed by the smart device 14.

[0403] [Second embodiment]

[0404] FIG. 3 shows an example of the configuration of a data processing system 210 according to the second embodiment.

[0405] 3, the data processing system 210 includes the data processing device 12 and smart glasses 214. An example of the data processing device 12 is a server.

[0406] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0407] The smart glasses 214 include a computer 36, a microphone 238, a speaker 240, a camera 42, and a communication I / F 44. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, and the camera 42 are also connected to the bus 52.

[0408] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0409] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0410] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0411] Fig. 4 shows an example of the main functions of the data processing device 12 and the smart glasses 214. As shown in Fig. 4, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0412] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0413] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0414] In the smart glasses 214, the reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0415] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the smart glasses 214 will be referred to as the "terminal."

[0416] This invention relates to a system for evaluating and examining AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[0417] Overall Overview

[0418] The server collects the necessary training data from the company's internal database or external API, normalizes it, detects personal information and trade secrets from the normalized data, and evaluates the data for bias. A report is generated based on the evaluation results, and the server notifies the user of the report.

[0419] Details of each step

[0420] Data collection

[0421] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[0422] Data normalization

[0423] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[0424] Personal information detection

[0425] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[0426] Trade Secret Detection

[0427] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[0428] Assessing data bias

[0429] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[0430] Generate reports

[0431] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias. This report includes the detected issues and how to address them. For example, the report is generated in PDF or HTML format and the contents are summarized.

[0432] User Notification

[0433] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[0434] Specific examples

[0435] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. A report is generated based on the evaluation results, and the report is notified to a person in charge within the company. The person in charge reviews the report and takes necessary measures, enabling the development of a reliable AI model.

[0436] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[0437] The processing flow will be explained below.

[0438] Step 1: Collect data

[0439] The server collects learning data from the company's internal databases and external APIs.

[0440] Action 1.1: Obtain database connection information and establish a connection.

[0441] Action 1.2: Execute a query on the data to be collected and retrieve the data. For example, query customer purchase history data.

[0442] Action 1.3: Save the acquired data to temporary storage.

[0443] Step 2: Normalize the data

[0444] The server normalizes the collected data.

[0445] Action 2.1: Standardize data formats, such as date formats and monetary values.

[0446] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[0447] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[0448] Step 3: Discover personal information

[0449] The server detects personal information from the normalized data.

[0450] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[0451] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[0452] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[0453] Step 4: Discover trade secrets

[0454] The server detects trade secrets contained in the data.

[0455] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[0456] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[0457] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[0458] Step 5: Assess data for bias

[0459] The server evaluates the collected data for bias.

[0460] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[0461] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[0462] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[0463] Step 6: Generate reports

[0464] The server generates a report based on the above evaluation results.

[0465] Action 6.1: Compile a list of discovered personal information and trade secrets.

[0466] Action 6.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[0467] Action 6.3: Write a report detailing the problem and its solution.

[0468] Action 6.4: Output the completed report in PDF or HTML format and save it on the server.

[0469] Step 7: Notify users

[0470] The server notifies the user of the generated report.

[0471] Action 7.1: Create a notification to let the user know where to save the report.

[0472] Action 7.2: Send a notification to the user via email or corporate notification system.

[0473] Action 7.3: Provide an interface for receiving feedback from the user.

[0474] Example 1

[0475] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0476] Conventional data collection and evaluation systems often failed to properly normalize the collected data, and it was difficult to detect personal or confidential information. Furthermore, evaluating whether the data was biased and generating reports based on the evaluation results was time-consuming and prone to human error. Furthermore, notifications of generated reports to users were often done manually, which was inefficient. A highly reliable system was needed to solve these problems.

[0477] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0478] In this invention, the server includes means for collecting information from a database or external API to obtain information to be collected, means for normalizing the collected information, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized information, means for detecting confidential information contained in the information, means for evaluating whether the collected information is biased, means for generating a report based on the results of the above evaluations, and means for notifying the user of the generated report.This enables efficient data collection, evaluation, reporting, and notification while increasing the accuracy of the data.

[0479] A "data collection target" is a set of information required for analysis or evaluation.

[0480] A "database" is a digital structure that organizes information and allows it to be searched and updated.

[0481] An "external API" is an interface for accessing external systems and services to obtain data.

[0482] "Information" means data or textual content that is collected, processed, and evaluated for a specific purpose or use.

[0483] "Normalization" is the process of standardizing data formats, removing unnecessary data, and filling in missing values.

[0484] A "missing value" is a value that is not recorded or is missing in a dataset.

[0485] "Personal information" means information that can be used to identify a specific individual, such as name, address, or telephone number.

[0486] "Confidential Information" is non-public, sensitive data that is specific to a particular organization or project.

[0487] "Bias" is a state in which data is overly concentrated or absent in a particular direction or attribute.

[0488] A "report" is a documented result of a particular analysis or evaluation, in a format that allows a user to review the contents.

[0489] "User" is a person or organization that receives reports generated by this system and takes action or makes decisions based on them.

[0490] This invention relates to a system for evaluating and examining AI learning data, and is implemented primarily using a server. Below, the program processing of this system is explained in natural language.

[0491] First, the server uses a database or external API to obtain the information to be collected. Specifically, data is collected using an SQL database (e.g., MySQL, PostgreSQL) or a NoSQL database (e.g., MongoDB). For example, a company's server collects customer purchase history data and temporarily stores it in storage (e.g., Amazon S3).

[0492] The server then normalizes the collected data. This process uses data processing libraries such as Apache Spark and Pandas. Data quality is improved by standardizing data formats, imputing missing values, and removing duplicate data. For example, standardizing date formats and replacing NULL values ​​with a mean value are performed.

[0493] Next, the server detects personal information from the normalized data using regular expressions (e.g., Python's re module) or machine learning models (e.g., spaCy, NLTK). For example, personal information such as name, address, phone number, and email address is detected and temporarily stored as a list.

[0494] Furthermore, the server detects confidential information contained in the data. It extracts confidential information using specific keyword lists and machine learning models (e.g., BERT, GPT). For example, project names and product development information are identified and stored as a list.

[0495] The server then assesses the collected data for bias using Matplotlib and statistical methods (e.g., t-tests, chi-squared tests), visually assessing data distribution with histograms and scatter plots, and performing statistical tests to determine bias.

[0496] Based on the results of the evaluation, the server generates a report in PDF (e.g., ReportLab) or HTML (e.g., Jinja2 template engine) format that includes the issues found and how to address them.

[0497] Finally, the server notifies the user of the generated report via email (e.g., SMTP server) or an internal corporate notification system (e.g., Slack, Microsoft Teams). For example, a link indicating where the generated report is saved may be sent via email.

[0498] Specific examples

[0499] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects purchase history data from a database (e.g., MySQL). Next, it normalizes the collected data using Pandas and detects personal or confidential information using regular expressions and spaCy. It then visualizes the data distribution using Matplotlib and performs statistical tests to evaluate data bias. Based on the evaluation results, it generates a report using ReportLab, and the report is sent to the responsible person via email via an SMTP server. The responsible person reviews the report and takes any necessary measures, enabling the development of a reliable AI model.

[0500] Prompt Sentence Examples

[0501] "Create a program to collect customer purchase history data, normalize the data, and detect personal and confidential information. Also, evaluate the data for bias, generate a PDF report, and notify the user of the report via email."

[0502] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[0503] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0504] Step 1:

[0505] The server collects information from a database or external API. It takes a query or API request as input and the collected data as output. For example, to get a user's purchasing history, it runs an SQL query to retrieve the information from the database. The server then temporarily stores this data in storage.

[0506] Step 2:

[0507] The server normalizes the collected data. It requires the collected raw data as input and obtains normalized data as output. Specific operations include standardizing data formats, imputing missing values, and removing duplicate data using Pandas. For example, it standardizes date formats and imputes NULL values ​​with the average value.

[0508] Step 3:

[0509] The server detects personal information from the normalized data. Normalized data is required as input, and the detected personal information is listed as output. Specifically, it uses regular expressions and spaCy to identify personal information such as name, address, phone number, and email address. This information is temporarily stored in storage.

[0510] Step 4:

[0511] The server detects confidential information contained in the data. Normalized data is required as input, and the detected confidential information is listed as output. Specifically, it uses BERT and a specific keyword list to extract confidential information such as project names and product development information. This information is also temporarily stored in storage.

[0512] Step 5:

[0513] The server evaluates whether the collected data is biased. Normalized data is required as input, and the output is the bias evaluation result. Specifically, Matplotlib is used to visually evaluate the data distribution using histograms and scatter plots, and statistical tests such as the chi-square test are performed. The evaluation results are compiled into a report.

[0514] Step 6:

[0515] The server generates a report based on the results of the assessment of detected personal information, confidential information, and bias. It takes as input the results of each detection and assessment, and as output the generated report. It uses ReportLab or the Jinja2 template engine to create reports in PDF and HTML format.

[0516] Step 7:

[0517] The server notifies the user of the generated report. The generated report and the user's contact information are required as input, and a notification is sent to the user as output. Specifically, it sends an email via an SMTP server and provides a link to the location where the report is saved. An in-house notification system can also be used as a notification method.

[0518] (Application example 1)

[0519] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0520] Sensor data and camera video data collected by autonomous vehicles may contain personal information and trade secrets, so systems are needed to properly process this data while ensuring data accuracy and compliance. Furthermore, if the collected data is biased, it can affect the learning results of the AI ​​model, reducing safety and reliability. Especially for autonomous vehicles, data bias can have a direct impact on the vehicle's driving behavior, so it is necessary to evaluate whether or not there is data bias and correct it appropriately.

[0521] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0522] In this invention, the server includes means for collecting data from a database or an external API to acquire data to be collected, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for collecting and normalizing sensor data and camera video data from an autonomous vehicle and evaluating whether there is bias in the personal information and data, means for generating a report based on the above evaluation results, and means for notifying a user of the generated report. This makes it possible to evaluate data bias and develop and operate a highly reliable AI model while ensuring data accuracy and compliance in autonomous vehicles.

[0523] A "database" is a set of structured data that allows for efficient storage, management, retrieval, and updating of information.

[0524] An "external API" is an interface for communicating with and exchanging data with other systems or services.

[0525] "Normalization" is the process of standardizing the format of data and imputing missing values.

[0526] "Personal information" is information that can identify a specific individual (e.g., name, address, telephone number, email address).

[0527] "Trade secrets" are information that is kept secret in order to maintain a company's competitive edge, and include development information for specific products and internal project names.

[0528] "Data bias" refers to a situation in which collected data is unevenly distributed across certain attributes (e.g., gender, age, region).

[0529] "Sensor data" refers to data collected from various sensors (e.g., GPS, accelerometer, camera, etc.).

[0530] "Camera video data" refers to still images and video data captured by a camera.

[0531] A "report" is a document that summarizes the results of a particular evaluation or analysis.

[0532] "User" means a person or entity that uses the System.

[0533] An "on-board computer" is a computing device installed in a vehicle and used to process data from sensors and cameras.

[0534] To implement the present invention, a system can be constructed in the following manner: The server performs processes for acquiring, normalizing, detecting, evaluating, generating reports, and notifying data to be collected.

[0535] Program processing and use of hardware and software

[0536] 1. Data Collection

[0537] The server uses a database and external API to collect sensor data and camera video data from the autonomous vehicle, allowing it to understand the vehicle's operating status in real time. For example, it uses the requests library to send HTTP requests and retrieve data.

[0538] 2. Data normalization

[0539] The server performs data normalization to standardize the collected data into a consistent format and impute missing values, using libraries such as pandas and sklearn.preprocessing, including standardizing date formats and scaling sensor data.

[0540] 3. Personal information detection

[0541] The server uses regular expressions and machine learning models to detect personal information from the normalized data, specifically using the re library to identify names, addresses, phone numbers, email addresses, etc.

[0542] 4. Trade Secret Discovery

[0543] The data collected by companies may contain trade secrets, which the server detects using a specific list of keywords and machine learning models.

[0544] 5. Assessing data bias

[0545] The server aggregates statistical information such as gender, age, and region to assess whether the collected data is biased, and performs statistical tests. The pandas library is used to generate histograms and scatter plots to visually evaluate the data distribution.

[0546] 6. Report Generation

[0547] Based on the results of the assessment of detected personal information, trade secrets, and data bias, the server generates a report in PDF or HTML format detailing the issues and how to address them.

[0548] 7. Notice to Users

[0549] Once a report is generated, the user is notified via email or the company's internal notification system, for example by sending a link to the location where the report is saved.

[0550] Specific examples

[0551] Below are some specific examples of data collection and analysis in autonomous vehicles.

[0552] 1. Example prompt:

[0553] The API collects sensor data and camera video data from an autonomous vehicle, retrieves the data every 10 seconds, and stores it on the vehicle's internal computer. The API endpoints are provided below (sensor: http: / / example.com / sensor_data, camera: http: / / example.com / camera_data).

[0554]

[0555] Scale the collected data and standardize the date format to 'YYYY-MM-DD HH:MM:SS'.

[0556]

[0557] Use regular expressions to detect and list names, phone numbers, and email addresses in your data.

[0558]

[0559] Use gender and age demographics to assess whether your dataset is biased.

[0560] In this way, the system can support the development and operation of reliable AI models while ensuring a high level of security and compliance for data collected from autonomous vehicles.

[0561] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0562] Step 1:

[0563] The server collects sensor data and camera video data from autonomous vehicles from databases and external APIs.

[0564] Input: Database or API endpoint.

[0565] Data processing: Data is retrieved using HTTP requests and temporarily stored in the server's storage. Specifically, data is retrieved in real time using the requests library.

[0566] Output: The raw data collected.

[0567] Step 2:

[0568] The server normalizes the collected data, standardizes it into a consistent format, and imputes missing values.

[0569] Input: Raw data collected.

[0570] Data processing: Using pandas and the sklearn.preprocessing library, we standardized date formats, scaled the data, and imputed missing values ​​with the mean and median.

[0571] Specifically, a data frame is created and the data format of each column is aligned.

[0572] Output: Normalized data.

[0573] Step 3:

[0574] The server detects personal information from the normalized data.

[0575] Input: Normalized data.

[0576] Data processing: Use the re library to extract names, addresses, phone numbers, and email addresses using regular expressions.

[0577] Specifically, pattern matching is performed to create a list of personal information.

[0578] Output: A list of personal information found.

[0579] Step 4:

[0580] The server detects trade secrets from the normalized data.

[0581] Input: Normalized data.

[0582] Data processing: Detecting specific information using company-defined keyword lists and machine learning models.

[0583] Specifically, it loads a machine learning model and makes predictions on the data.

[0584] Output: A list of discovered trade secrets.

[0585] Step 5:

[0586] The server evaluates the collected data for bias.

[0587] Input: Normalized data.

[0588] Data processing: Use the pandas library to aggregate statistics and evaluate the distribution of attribute values ​​such as gender and age.

[0589] Specifically, histograms and scatter plots are generated and biases are detected using statistical methods.

[0590] Output: Report of bias assessment.

[0591] Step 6:

[0592] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias.

[0593] Input: List of personal information, list of trade secrets, bias assessment report.

[0594] Data processing: Compile this information into a document in PDF or HTML format.

[0595] Specifically, data is inserted into an automatically generated template to generate a document.

[0596] Output: Compliance report.

[0597] Step 7:

[0598] The server notifies the user of the generated report.

[0599] Input: Compliance report.

[0600] Data processing: Send a link to the report to users via email or internal notification systems.

[0601] Specifically, it uses an SMTP server to send an email and includes a link.

[0602] Output: Notification email to user.

[0603] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0604] This invention combines an emotion engine with a system that evaluates and examines AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[0605] Overall Overview

[0606] The server collects the necessary learning data from internal company databases and external APIs and normalizes it. It then detects personal information and trade secrets from the normalized data and evaluates the data for bias. It generates a report based on the evaluation results and notifies the user of the report. The server also incorporates an emotion engine that recognizes user emotions based on the collected data and analyzes user feedback and responses.

[0607] Details of each step

[0608] Data collection

[0609] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[0610] Data normalization

[0611] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[0612] Personal information detection

[0613] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[0614] Trade Secret Detection

[0615] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[0616] Assessing data bias

[0617] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[0618] Emotion recognition by emotion engine

[0619] The server recognizes the user's emotions based on the collected data. This emotion recognition involves analyzing voice data and text data. For example, the server analyzes voice data to identify the user's emotions (happiness, sadness, anger, etc.). It also analyzes text data using natural language processing (NLP) techniques to identify the user's emotions.

[0620] Generate reports

[0621] The server generates a report based on the results of the user sentiment analysis, as well as the results of the evaluation of detected personal information and trade secrets, and data bias. This report includes detected problems and how to address them, as well as user feedback based on the sentiment analysis. For example, the report is generated in PDF or HTML format and the contents are summarized.

[0622] User Notification

[0623] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[0624] Specific examples

[0625] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the content. Finally, this report is notified to a person in the company who reviews it and takes necessary measures, enabling the development of a reliable AI model.

[0626] This system allows companies to identify potential risks in AI learning data in advance and take appropriate measures, and by taking user emotions into consideration, it is possible to obtain more accurate analysis results. This makes it possible to develop highly reliable AI and significantly reduce operational risks.

[0627] The processing flow will be explained below.

[0628] Step 1: Collect data

[0629] The server collects learning data from the company's internal databases and external APIs.

[0630] Action 1.1: Obtain database connection information and establish a connection.

[0631] Action 1.2: Query the data to be collected and retrieve the data. For example, query customer purchase history data.

[0632] Action 1.3: Save the acquired data to temporary storage.

[0633] Step 2: Normalize the data

[0634] The server normalizes the collected data.

[0635] Action 2.1: Standardize data formats, such as date formats and monetary values.

[0636] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[0637] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[0638] Step 3: Discover personal information

[0639] The server detects personal information from the normalized data.

[0640] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[0641] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[0642] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[0643] Step 4: Discover trade secrets

[0644] The server detects trade secrets contained in the data.

[0645] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[0646] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[0647] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[0648] Step 5: Assess data for bias

[0649] The server evaluates the collected data for bias.

[0650] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[0651] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[0652] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[0653] Step 6: Emotion Recognition with the Emotion Engine

[0654] The server recognizes the user's emotions based on the collected data.

[0655] Action 6.1: Analyze audio data. For example, take an audio file and identify emotions such as joy, sadness, or anger from the audio.

[0656] Activity 6.2: Analyze text data, for example customer reviews and feedback, using natural language processing techniques to identify sentiment.

[0657] Action 6.3: List the sentiment analysis results and save them in temporary storage.

[0658] Step 7: Generate reports

[0659] The server generates a report based on the above evaluation results.

[0660] Action 7.1: Compile a list of discovered personal information and trade secrets.

[0661] Action 7.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[0662] Action 7.3: Summarize user feedback based on sentiment analysis results.

[0663] Action 7.4: Output the completed report in PDF or HTML format and save it on the server.

[0664] Step 8: Notify users

[0665] The server notifies the user of the generated report.

[0666] Action 8.1: Create a notification to let the user know where the report is saved.

[0667] Action 8.2: Send a notification to the user via email or corporate notification system.

[0668] Action 8.3: Provide an interface for receiving feedback from the user.

[0669] Example 2

[0670] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0671] Conventional data collection and analysis systems have had problems such as difficulty in identifying personal information or trade secrets contained in the collected training data, and insufficient evaluation of data bias. Furthermore, because data is evaluated without taking user emotions into consideration, the accuracy of the final analysis results is likely to be reduced.

[0672] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[0673] In this invention, the server includes means for collecting data from databases and external information sources, means for normalizing the collected data, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for analyzing voice data and text data to recognize user emotions, means for generating a report based on the above evaluation results and emotion recognition results, and means for notifying the user of the generated report. This makes it possible to identify potential risks in the data in advance, take appropriate measures, and perform highly accurate analysis that takes user emotions into consideration.

[0674] A "database" is a computer system that systematically stores information and allows it to be efficiently searched, manipulated, and updated.

[0675] An "external source" is a data source that resides outside the enterprise and is typically accessed via an API.

[0676] "Normalization" is the process of standardizing the format of collected data and filling in missing values.

[0677] "Personal information" is information that can identify a specific individual, and includes, for example, name, address, telephone number, and email address.

[0678] "Trade secrets" are important, non-public information about a company's business, including, for example, information about new products under development and details of internal projects.

[0679] "Data bias" refers to imbalances in a dataset based on specific attributes or conditions.

[0680] "Emotion recognition" is the process of analyzing voice and text data to identify a user's emotions (e.g., joy, sadness, anger, etc.).

[0681] A "report" is a document summarizing detected personal information, trade secrets, data bias, and emotion recognition results.

[0682] "Notification" refers to the act of informing the user of the contents of the generated report, and is carried out via email or an in-company notification system.

[0683] This invention relates to a system that automatically performs processes from data collection to analysis and report generation. This system will be described in detail below.

[0684] Data collection

[0685] The server has a means for collecting training data from the company's internal database and external information sources. For example, the server executes SQL queries against the database to obtain purchase history data. It also obtains data from external information sources using HTTP requests and performs authentication using an API key.

[0686] For example, to retrieve purchase history from the database, the server executes the following SQL query:

[0687] SELECT FROM PurchaseHistory WHERE Date >= '2023-01-01'

[0688] To collect data from an external API, send an HTTP request like this:

[0689] GET https: / / api.example.com / reviews?product_id=123

[0690] Data normalization

[0691] The server normalizes the collected data by standardizing the data format and imputing missing values, such as standardizing date formats to "YYYY-MM-DD," standardizing numeric display formats, and replacing NULL values ​​with the mean or median.

[0692] For example, to standardize date formats, do the following:

[0693] Convert 2023 / 01 / 01 to 2023-01-01

[0694] To fill in missing values, for example, if "purchase amount" is NULL, the overall average purchase amount is substituted.

[0695] Personal information and trade secret detection

[0696] The server uses regular expressions and machine learning models to detect personal information in the normalized data. For example, the regular expression to identify email addresses:

[0697] [a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,4}

[0698] Trade secret detection uses specific keyword lists and machine learning models, for example, scanning data for keywords such as "new product" or "project X."

[0699] Assessing data bias

[0700] The server evaluates the collected data for bias. This evaluation process involves aggregating demographic information such as age, gender, and region, and performing statistical tests. For example, chi-square tests and t-tests are used to verify statistically significant bias. This process can use Python's Matplotlib library to generate histograms and scatter plots.

[0701] emotion recognition

[0702] The server analyzes the voice and text data to recognize the user's emotions. This emotion recognition uses a speech recognition engine (such as the Google Cloud Speech-to-Text API) to convert voice data into text, and natural language processing (NLP) technology to analyze text data.

[0703] As a concrete example, we use Waveform Transform and Google Cloud Speech-to-Text API to analyze audio files and identify user emotions.

[0704] Report generation and notification

[0705] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. This report is generated in PDF or HTML format and saved on the server. For example, the Python ReportLab library is used to generate the PDF report.

[0706] The generated report will be sent to the user via email or the company's internal notification system, for example, a link to the location where the generated report is saved will be sent via email:

[0707] http: / / example.com / reports / Report2023-10-01.pdf

[0708] Specific examples

[0709] For example, suppose a company wants to develop a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Then, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the results. Finally, the report is sent to a person in the company who reviews it and takes any necessary measures.

[0710] Prompt Sentence Examples

[0711] For example, here is an example of a prompt sentence being input to a generative AI model:

[0712] I want to create an AI model that uses customer purchase history data to infer user preferences and interests and propose personalized marketing strategies. How can I build a system that collects data, normalizes it, detects personal information and trade secrets, evaluates data bias, recognizes emotions using an emotion engine, generates reports, and notifies users?

[0713] The flow of the identification process in the second embodiment will be described with reference to FIG.

[0714] Step 1: Collect data

[0715] The server collects data from a company's internal database or external information source. The server first connects to the database and uses authentication information to confirm access rights. For example, it executes an SQL query to retrieve purchase history data. The input is the database connection information and SQL query, and the output is a record of purchase history data. When collecting data from an external information source, the server authenticates using an API key and sends an API request. Specifically, the server sends an HTTP request and temporarily stores the data obtained as a response in storage on the server.

[0716] Step 2: Normalize the data

[0717] The server normalizes the collected data. In this process, the data format is standardized and missing values ​​are filled. For example, the date format is standardized to "YYYY-MM-DD" and NULL values ​​are replaced with the average or median. The input is the collected purchase history data, and the output is the data converted into a standardized format. Specifically, the server applies the rules for data format conversion and missing value filling, and temporarily stores the processed data.

[0718] Step 3: Discover personal information

[0719] The server detects personal information from the normalized data. Regular expressions and machine learning models are used for detection. The input is the normalized data, and the output is a list of detected personal information. Specifically, the server uses predefined regular expressions to identify email addresses and phone numbers, and uses machine learning models to identify addresses and names. The detection results are compiled into a list and stored in temporary storage.

[0720] Step 4: Discover trade secrets

[0721] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for a specific product or the names of internal projects. The input is normalized data, and the output is a list of detected trade secrets. Specifically, the server applies a keyword matching algorithm to extract relevant information. The detection results are listed and temporarily stored.

[0722] Step 5: Assess data for bias

[0723] The server evaluates whether the collected data is biased. This evaluation process involves compiling statistical information such as age, gender, and region, and performing statistical tests. The input is normalized data, and the output is a statistical report indicating whether bias is present or absent. Specifically, the server uses Python libraries to generate histograms and scatter plots and perform chi-square tests and t-tests. This allows the distribution of data to be visualized and statistical bias to be evaluated.

[0724] Step 6: Emotion Recognition

[0725] The server analyzes the voice and text data to recognize the user's emotions. Emotion recognition uses a voice recognition engine and natural language processing (NLP) technology. The input is voice and text data, and the output is analyzed emotion data. Specifically, the server converts the voice data into text (using the Google Cloud Speech-to-Text API) and analyzes the text using a sentiment analysis model. The analysis results are stored in a database.

[0726] Step 7: Generate reports

[0727] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. The input is all of these detection and assessment results, and the output is a report in PDF or HTML format. Specifically, the server uses the Python ReportLab library to generate a PDF report and summarize the report contents. The generated report is saved on the server.

[0728] Step 8: Notify users

[0729] The server notifies the user of the generated report. The input is the generated report and the user's contact information, and the output is a confirmation that the user has been notified. Specifically, the server retrieves the user's email address from a database, generates an email, and sends it using an SMTP server. The email includes a link indicating where the report is saved.

[0730] (Application example 2)

[0731] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the smart glasses 214 will be referred to as a "terminal."

[0732] In autonomous vehicles, in addition to analyzing driving data and evaluating bias, it is necessary to properly analyze the emotional feedback of the user while driving in order to optimize safety and operation management. Conventional systems lack the functionality to properly evaluate and process this emotional data, which has led to problems such as being unable to adequately evaluate bias in driving data and user stress.

[0733] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[0734] In this invention, the server includes means for collecting data from databases and external APIs, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for incorporating an emotion engine that recognizes user emotions, means for generating a report based on the evaluation results and emotion recognition results, and means for notifying the user of the generated report. This enables comprehensive analysis of autonomous vehicle driving data and user emotional feedback, enabling optimization of safety and operation management.

[0735] A "database" is an organized collection of data, a system that stores various data and allows it to be easily searched, retrieved, and analyzed.

[0736] An "external API" is a program interface provided for linking with other software or systems, and is a means of obtaining and manipulating data.

[0737] "Normalization" is the process of standardizing the format of collected data and imputing missing values, improving data quality and consistency.

[0738] "Personal information" is information that identifies a specific individual, such as name, address, and telephone number.

[0739] A "trade secret" is non-public information held by a company or organization, such as technical or commercial information that is kept secret in order to maintain a competitive advantage.

[0740] Bias assessment is the process of detecting and analyzing biases and imbalances present in collected data.

[0741] An "emotion engine" is a technology for recognizing and analyzing a user's emotions, and is a system that estimates their emotional state from voice and text data.

[0742] A "report" is a report summarizing the results of analysis and evaluation of collected data, and is provided to users.

[0743] "Notification" is the process of communicating generated reports and important information to users, typically via email or smartphone notification systems.

[0744] This invention relates to a system for comprehensively analyzing driving data of autonomous vehicles and emotional feedback from users to optimize safety and operation management. The system includes a server, data collection means, data normalization means, personal information detection means, trade secret detection means, bias evaluation means, an emotion recognition engine, report generation means, and user notification means.

[0745] Server Configuration

[0746] The server collects various sensor data (e.g., GPS, speed, battery status, etc.) sent from the autonomous vehicle via a database or external API. The collected data is temporarily stored on the server. Data collection methods include scripts using programming languages ​​such as Python and libraries that call external APIs. For example, the requests library can be used to obtain the required data from external APIs.

[0747] Data normalization measures

[0748] Data normalization involves standardizing the format of the collected data and imputing missing values. To do this, we use the pandas library to manipulate data frames and perform the necessary format conversions and imputation of missing values. For example, we standardize date formats and replace NULL values ​​with the average value.

[0749] Personal information detection method

[0750] Personal information detection techniques identify personal information from normalized data using regular expressions and machine learning models. For example, email addresses and phone numbers can be extracted using regular expressions, and names and addresses can be detected using machine learning models (e.g., models using the scikit-learn library).

[0751] Trade secret detection methods

[0752] The trade secret detection method detects trade secrets in data. It uses a keyword list specified by the company and a machine learning model to detect internal project names and technical information. This also uses the pandas library and natural language processing techniques (e.g., the nltk library and spaCy library).

[0753] Bias assessment measures

[0754] The bias assessment tool aggregates statistical information from the collected data to detect abnormal bias. For example, it creates histograms and scatter plots using the matplotlib and seaborn libraries and performs statistical tests.

[0755] Emotion Recognition Engine

[0756] The emotion recognition engine analyzes voice data and text feedback while driving to identify the user's emotions. To analyze voice data, it uses a voice recognition library (e.g., the SpeechRecognition library), and to analyze the emotions of text data, it uses the TextBlob or VADER library.

[0757] Report Generation Method

[0758] The report generation tool generates reports based on detected personal information, trade secrets, data bias, and sentiment analysis results. These reports are generated in PDF and HTML formats, and are visually easy to understand by utilizing the matplotlib and pdfkit libraries.

[0759] User notification method

[0760] The user notification means notifies the user of the generated report via email or in-app notification. Email is sent using the smtplib library, and a push notification library (e.g., Pushy) is used for in-app notifications.

[0761] Specific examples

[0762] For example, if an autonomous vehicle senses stress due to traffic conditions and finds bias in its driving data, the system will detect the bias and stress in the driving data and provide a detailed report and improvement measures to the operations manager.

[0763] Prompt Sentence Examples

[0764] Please explain in natural language the processes of the "Self-Driving Vehicle Emotion Feedback & Data Analysis System": [Self-Driving Vehicle Sensor Data], [Driving Data Normalization], [Personal Information Anonymization], [Data Bias Assessment], [Emotion Recognition], [Report Generation], and [User Notification].

[0765] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[0766] Step 1:

[0767] The server collects sensor data from autonomous vehicles. Input data includes GPS data, speed data, battery data, etc., which are obtained from databases and external APIs. Specifically, the server makes API calls, retrieves data using the requests library, and temporarily stores it. The output is the collected raw data.

[0768] Step 2:

[0769] The server normalizes the data collected in step 1. The input data is raw data, which is stored in a data frame using the PANDAS library. Specific operations include standardizing date formats, converting NULL values ​​to averages, and removing duplicate data. The output is normalized data.

[0770] Step 3:

[0771] The server detects personal information from the normalized data. The input data is normalized data, and regular expressions and machine learning models from the scikit-learn library are used. Specifically, it identifies names, addresses, and phone numbers, creates a list, and saves it in temporary storage. The output is a list of detected personal information.

[0772] Step 4:

[0773] The server detects trade secrets contained in the data. The input data is normalized data, and uses a keyword list specified by the company and natural language processing technology (nltk library and spaCy library). Specifically, it detects specific products and project names, creates a list, and saves it in temporary storage. The output is a list of detected trade secrets.

[0774] Step 5:

[0775] The server evaluates the bias of the collected data. The input data is normalized data, and it is visualized using the matplotlib and seaborn libraries and subjected to statistical testing. Specific operations include creating histograms and scatter plots and determining statistical bias. The output is the bias evaluation results.

[0776] Step 6:

[0777] The server incorporates an emotion engine that recognizes the user's emotions. The input data is voice data and text data from the driver's seat, and it uses a voice recognition library, TextBlob, and VADER library. Specific operations include analyzing voice data, identifying emotions, and identifying emotions from text data. The output is the emotion analysis results.

[0778] Step 7:

[0779] The server generates a report based on the detected personal information, trade secrets, data bias evaluation results, and sentiment analysis results. The input data are the evaluation results obtained at each step, and it uses the matplotlib and pdfkit libraries. Specific operations include visualizing the evaluation results and generating reports in PDF and HTML format. The output is the generated report.

[0780] Step 8:

[0781] The server notifies the user of the generated report. The input data is the generated report, and the smtplib library is used to send emails or push notifications. Specific operations include creating a notification message and sending an email or in-app notification. The output indicates that the notification has been completed.

[0782] The specific processing unit 290 transmits the result of the specific processing to the smart glasses 214. In the smart glasses 214, the control unit 46A causes the speaker 240 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[0783] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[0784] In the above embodiment, an example in which the specific processing is performed by the data processing device 12 has been given, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the smart glasses 214.

[0785] [Third embodiment]

[0786] FIG. 5 shows an example of the configuration of a data processing system 310 according to the third embodiment.

[0787] 5, the data processing system 310 includes the data processing device 12 and a headset type terminal 314. An example of the data processing device 12 is a server.

[0788] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[0789] The headset type terminal 314 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a display 343. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the display 343 are also connected to the bus 52.

[0790] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[0791] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[0792] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[0793] Fig. 6 shows an example of the main functions of the data processing device 12 and the headset type terminal 314. As shown in Fig. 6, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[0794] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[0795] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[0796] In the headset type terminal 314, a reception output process is performed by the processor 46. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[0797] Next, a description will be given of the identification process performed by the identification processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as the "server" and the headset type terminal 314 will be referred to as the "terminal."

[0798] This invention relates to a system for evaluating and examining AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[0799] Overall Overview

[0800] The server collects the necessary training data from the company's internal database or external API, normalizes it, detects personal information and trade secrets from the normalized data, and evaluates the data for bias. A report is generated based on the evaluation results, and the server notifies the user of the report.

[0801] Details of each step

[0802] Data collection

[0803] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[0804] Data normalization

[0805] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[0806] Personal information detection

[0807] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[0808] Trade Secret Detection

[0809] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[0810] Assessing data bias

[0811] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[0812] Generate reports

[0813] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias. This report includes the detected issues and how to address them. For example, the report is generated in PDF or HTML format and the contents are summarized.

[0814] User Notification

[0815] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[0816] Specific examples

[0817] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. A report is generated based on the evaluation results, and the report is notified to a person in charge within the company. The person in charge reviews the report and takes necessary measures, enabling the development of a reliable AI model.

[0818] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[0819] The processing flow will be explained below.

[0820] Step 1: Collect data

[0821] The server collects learning data from the company's internal databases and external APIs.

[0822] Action 1.1: Obtain database connection information and establish a connection.

[0823] Action 1.2: Execute a query on the data to be collected and retrieve the data. For example, query customer purchase history data.

[0824] Action 1.3: Save the acquired data to temporary storage.

[0825] Step 2: Normalize the data

[0826] The server normalizes the collected data.

[0827] Action 2.1: Standardize data formats, such as date formats and monetary values.

[0828] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[0829] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[0830] Step 3: Discover personal information

[0831] The server detects personal information from the normalized data.

[0832] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[0833] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[0834] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[0835] Step 4: Discover trade secrets

[0836] The server detects trade secrets contained in the data.

[0837] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[0838] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[0839] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[0840] Step 5: Assess data for bias

[0841] The server evaluates the collected data for bias.

[0842] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[0843] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[0844] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[0845] Step 6: Generate reports

[0846] The server generates a report based on the above evaluation results.

[0847] Action 6.1: Compile a list of discovered personal information and trade secrets.

[0848] Action 6.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[0849] Action 6.3: Write a report detailing the problem and its solution.

[0850] Action 6.4: Output the completed report in PDF or HTML format and save it on the server.

[0851] Step 7: Notify users

[0852] The server notifies the user of the generated report.

[0853] Action 7.1: Create a notification to let the user know where to save the report.

[0854] Action 7.2: Send a notification to the user via email or corporate notification system.

[0855] Action 7.3: Provide an interface for receiving feedback from the user.

[0856] Example 1

[0857] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0858] Conventional data collection and evaluation systems often failed to properly normalize the collected data, and it was difficult to detect personal or confidential information. Furthermore, evaluating whether the data was biased and generating reports based on the evaluation results was time-consuming and prone to human error. Furthermore, notifications of generated reports to users were often done manually, which was inefficient. A highly reliable system was needed to solve these problems.

[0859] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[0860] In this invention, the server includes means for collecting information from a database or external API to obtain information to be collected, means for normalizing the collected information, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized information, means for detecting confidential information contained in the information, means for evaluating whether the collected information is biased, means for generating a report based on the results of the above evaluations, and means for notifying the user of the generated report.This enables efficient data collection, evaluation, reporting, and notification while increasing the accuracy of the data.

[0861] A "data collection target" is a set of information required for analysis or evaluation.

[0862] A "database" is a digital structure that organizes information and allows it to be searched and updated.

[0863] An "external API" is an interface for accessing external systems and services to obtain data.

[0864] "Information" means data or textual content that is collected, processed, and evaluated for a specific purpose or use.

[0865] "Normalization" is the process of standardizing data formats, removing unnecessary data, and filling in missing values.

[0866] A "missing value" is a value that is not recorded or is missing in a dataset.

[0867] "Personal information" means information that can be used to identify a specific individual, such as name, address, or telephone number.

[0868] "Confidential Information" is non-public, sensitive data that is specific to a particular organization or project.

[0869] "Bias" is a state in which data is overly concentrated or absent in a particular direction or attribute.

[0870] A "report" is a documented result of a particular analysis or evaluation, in a format that allows a user to review the contents.

[0871] "User" is a person or organization that receives reports generated by this system and takes action or makes decisions based on them.

[0872] This invention relates to a system for evaluating and examining AI learning data, and is implemented primarily using a server. Below, the program processing of this system is explained in natural language.

[0873] First, the server uses a database or external API to obtain the information to be collected. Specifically, data is collected using an SQL database (e.g., MySQL, PostgreSQL) or a NoSQL database (e.g., MongoDB). For example, a company's server collects customer purchase history data and temporarily stores it in storage (e.g., Amazon S3).

[0874] The server then normalizes the collected data. This process uses data processing libraries such as Apache Spark and Pandas. Data quality is improved by standardizing data formats, imputing missing values, and removing duplicate data. For example, standardizing date formats and replacing NULL values ​​with a mean value are performed.

[0875] Next, the server detects personal information from the normalized data using regular expressions (e.g., Python's re module) or machine learning models (e.g., spaCy, NLTK). For example, personal information such as name, address, phone number, and email address is detected and temporarily stored as a list.

[0876] Furthermore, the server detects confidential information contained in the data. It extracts confidential information using specific keyword lists and machine learning models (e.g., BERT, GPT). For example, project names and product development information are identified and stored as a list.

[0877] The server then assesses the collected data for bias using Matplotlib and statistical methods (e.g., t-tests, chi-squared tests), visually assessing data distribution with histograms and scatter plots, and performing statistical tests to determine bias.

[0878] Based on the results of the evaluation, the server generates a report in PDF (e.g., ReportLab) or HTML (e.g., Jinja2 template engine) format that includes the issues found and how to address them.

[0879] Finally, the server notifies the user of the generated report via email (e.g., SMTP server) or an internal corporate notification system (e.g., Slack, Microsoft Teams). For example, a link indicating where the generated report is saved may be sent via email.

[0880] Specific examples

[0881] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects purchase history data from a database (e.g., MySQL). Next, it normalizes the collected data using Pandas and detects personal or confidential information using regular expressions and spaCy. It then visualizes the data distribution using Matplotlib and performs statistical tests to evaluate data bias. Based on the evaluation results, it generates a report using ReportLab, and the report is sent to the responsible person via email via an SMTP server. The responsible person reviews the report and takes any necessary measures, enabling the development of a reliable AI model.

[0882] Prompt Sentence Examples

[0883] "Create a program to collect customer purchase history data, normalize the data, and detect personal and confidential information. Also, evaluate the data for bias, generate a PDF report, and notify the user of the report via email."

[0884] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[0885] The flow of the identification process in the first embodiment will be described with reference to FIG.

[0886] Step 1:

[0887] The server collects information from a database or external API. It takes a query or API request as input and the collected data as output. For example, to get a user's purchasing history, it runs an SQL query to retrieve the information from the database. The server then temporarily stores this data in storage.

[0888] Step 2:

[0889] The server normalizes the collected data. It requires the collected raw data as input and obtains normalized data as output. Specific operations include standardizing data formats, imputing missing values, and removing duplicate data using Pandas. For example, it standardizes date formats and imputes NULL values ​​with the average value.

[0890] Step 3:

[0891] The server detects personal information from the normalized data. Normalized data is required as input, and the detected personal information is listed as output. Specifically, it uses regular expressions and spaCy to identify personal information such as name, address, phone number, and email address. This information is temporarily stored in storage.

[0892] Step 4:

[0893] The server detects confidential information contained in the data. Normalized data is required as input, and the detected confidential information is listed as output. Specifically, it uses BERT and a specific keyword list to extract confidential information such as project names and product development information. This information is also temporarily stored in storage.

[0894] Step 5:

[0895] The server evaluates whether the collected data is biased. Normalized data is required as input, and the output is the bias evaluation result. Specifically, Matplotlib is used to visually evaluate the data distribution using histograms and scatter plots, and statistical tests such as the chi-square test are performed. The evaluation results are compiled into a report.

[0896] Step 6:

[0897] The server generates a report based on the results of the assessment of detected personal information, confidential information, and bias. It takes as input the results of each detection and assessment, and as output the generated report. It uses ReportLab or the Jinja2 template engine to create reports in PDF and HTML format.

[0898] Step 7:

[0899] The server notifies the user of the generated report. The generated report and the user's contact information are required as input, and a notification is sent to the user as output. Specifically, it sends an email via an SMTP server and provides a link to the location where the report is saved. An in-house notification system can also be used as a notification method.

[0900] (Application example 1)

[0901] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[0902] Sensor data and camera video data collected by autonomous vehicles may contain personal information and trade secrets, so systems are needed to properly process this data while ensuring data accuracy and compliance. Furthermore, if the collected data is biased, it can affect the learning results of the AI ​​model, reducing safety and reliability. Especially for autonomous vehicles, data bias can have a direct impact on the vehicle's driving behavior, so it is necessary to evaluate whether or not there is data bias and correct it appropriately.

[0903] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[0904] In this invention, the server includes means for collecting data from a database or an external API to acquire data to be collected, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for collecting and normalizing sensor data and camera video data from an autonomous vehicle and evaluating whether there is bias in the personal information and data, means for generating a report based on the above evaluation results, and means for notifying a user of the generated report. This makes it possible to evaluate data bias and develop and operate a highly reliable AI model while ensuring data accuracy and compliance in autonomous vehicles.

[0905] A "database" is a set of structured data that allows for efficient storage, management, retrieval, and updating of information.

[0906] An "external API" is an interface for communicating with and exchanging data with other systems or services.

[0907] "Normalization" is the process of standardizing the format of data and imputing missing values.

[0908] "Personal information" is information that can identify a specific individual (e.g., name, address, telephone number, email address).

[0909] "Trade secrets" are information that is kept secret in order to maintain a company's competitive edge, and include development information for specific products and internal project names.

[0910] "Data bias" refers to a situation in which collected data is unevenly distributed across certain attributes (e.g., gender, age, region).

[0911] "Sensor data" refers to data collected from various sensors (e.g., GPS, accelerometer, camera, etc.).

[0912] "Camera video data" refers to still images and video data captured by a camera.

[0913] A "report" is a document that summarizes the results of a particular evaluation or analysis.

[0914] "User" means a person or entity that uses the System.

[0915] An "on-board computer" is a computing device installed in a vehicle and used to process data from sensors and cameras.

[0916] To implement the present invention, a system can be constructed in the following manner: The server performs processes for acquiring, normalizing, detecting, evaluating, generating reports, and notifying data to be collected.

[0917] Program processing and use of hardware and software

[0918] 1. Data Collection

[0919] The server uses a database and external API to collect sensor data and camera video data from the autonomous vehicle, allowing it to understand the vehicle's operating status in real time. For example, it uses the requests library to send HTTP requests and retrieve data.

[0920] 2. Data normalization

[0921] The server performs data normalization to standardize the collected data into a consistent format and impute missing values, using libraries such as pandas and sklearn.preprocessing, including standardizing date formats and scaling sensor data.

[0922] 3. Personal information detection

[0923] The server uses regular expressions and machine learning models to detect personal information from the normalized data, specifically using the re library to identify names, addresses, phone numbers, email addresses, etc.

[0924] 4. Trade Secret Discovery

[0925] The data collected by companies may contain trade secrets, which the server detects using a specific list of keywords and machine learning models.

[0926] 5. Assessing data bias

[0927] The server aggregates statistical information such as gender, age, and region to assess whether the collected data is biased, and performs statistical tests. The pandas library is used to generate histograms and scatter plots to visually evaluate the data distribution.

[0928] 6. Report Generation

[0929] Based on the results of the assessment of detected personal information, trade secrets, and data bias, the server generates a report in PDF or HTML format detailing the issues and how to address them.

[0930] 7. Notice to Users

[0931] Once a report is generated, the user is notified via email or the company's internal notification system, for example by sending a link to the location where the report is saved.

[0932] Specific examples

[0933] Below are some specific examples of data collection and analysis in autonomous vehicles.

[0934] 1. Example prompt:

[0935] The API collects sensor data and camera video data from an autonomous vehicle, retrieves the data every 10 seconds, and stores it on the vehicle's internal computer. The API endpoints are provided below (sensor: http: / / example.com / sensor_data, camera: http: / / example.com / camera_data).

[0936]

[0937] Scale the collected data and standardize the date format to 'YYYY-MM-DD HH:MM:SS'.

[0938]

[0939] Use regular expressions to detect and list names, phone numbers, and email addresses in your data.

[0940]

[0941] Use gender and age demographics to assess whether your dataset is biased.

[0942] In this way, the system can support the development and operation of reliable AI models while ensuring a high level of security and compliance for data collected from autonomous vehicles.

[0943] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[0944] Step 1:

[0945] The server collects sensor data and camera video data from autonomous vehicles from databases and external APIs.

[0946] Input: Database or API endpoint.

[0947] Data processing: Data is retrieved using HTTP requests and temporarily stored in the server's storage. Specifically, data is retrieved in real time using the requests library.

[0948] Output: The raw data collected.

[0949] Step 2:

[0950] The server normalizes the collected data, standardizes it into a consistent format, and imputes missing values.

[0951] Input: Raw data collected.

[0952] Data processing: Using pandas and the sklearn.preprocessing library, we standardized date formats, scaled the data, and imputed missing values ​​with the mean and median.

[0953] Specifically, a data frame is created and the data format of each column is aligned.

[0954] Output: Normalized data.

[0955] Step 3:

[0956] The server detects personal information from the normalized data.

[0957] Input: Normalized data.

[0958] Data processing: Use the re library to extract names, addresses, phone numbers, and email addresses using regular expressions.

[0959] Specifically, pattern matching is performed to create a list of personal information.

[0960] Output: A list of personal information found.

[0961] Step 4:

[0962] The server detects trade secrets from the normalized data.

[0963] Input: Normalized data.

[0964] Data processing: Detecting specific information using company-defined keyword lists and machine learning models.

[0965] Specifically, it loads a machine learning model and makes predictions on the data.

[0966] Output: A list of discovered trade secrets.

[0967] Step 5:

[0968] The server evaluates the collected data for bias.

[0969] Input: Normalized data.

[0970] Data processing: Use the pandas library to aggregate statistics and evaluate the distribution of attribute values ​​such as gender and age.

[0971] Specifically, histograms and scatter plots are generated and biases are detected using statistical methods.

[0972] Output: Report of bias assessment.

[0973] Step 6:

[0974] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias.

[0975] Input: List of personal information, list of trade secrets, bias assessment report.

[0976] Data processing: Compile this information into a document in PDF or HTML format.

[0977] Specifically, data is inserted into an automatically generated template to generate a document.

[0978] Output: Compliance report.

[0979] Step 7:

[0980] The server notifies the user of the generated report.

[0981] Input: Compliance report.

[0982] Data processing: Send a link to the report to users via email or internal notification systems.

[0983] Specifically, it uses an SMTP server to send an email and includes a link.

[0984] Output: Notification email to user.

[0985] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[0986] This invention combines an emotion engine with a system that evaluates and examines AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[0987] Overall Overview

[0988] The server collects the necessary learning data from internal company databases and external APIs and normalizes it. It then detects personal information and trade secrets from the normalized data and evaluates the data for bias. It generates a report based on the evaluation results and notifies the user of the report. The server also incorporates an emotion engine that recognizes user emotions based on the collected data and analyzes user feedback and responses.

[0989] Details of each step

[0990] Data collection

[0991] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[0992] Data normalization

[0993] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[0994] Personal information detection

[0995] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[0996] Trade Secret Detection

[0997] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[0998] Assessing data bias

[0999] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[1000] Emotion recognition by emotion engine

[1001] The server recognizes the user's emotions based on the collected data. This emotion recognition involves analyzing voice data and text data. For example, the server analyzes voice data to identify the user's emotions (happiness, sadness, anger, etc.). It also analyzes text data using natural language processing (NLP) techniques to identify the user's emotions.

[1002] Generate reports

[1003] The server generates a report based on the results of the user sentiment analysis, as well as the results of the evaluation of detected personal information and trade secrets, and data bias. This report includes detected problems and how to address them, as well as user feedback based on the sentiment analysis. For example, the report is generated in PDF or HTML format and the contents are summarized.

[1004] User Notification

[1005] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[1006] Specific examples

[1007] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the content. Finally, this report is notified to a person in the company who reviews it and takes necessary measures, enabling the development of a reliable AI model.

[1008] This system allows companies to identify potential risks in AI learning data in advance and take appropriate measures, and by taking user emotions into consideration, it is possible to obtain more accurate analysis results. This makes it possible to develop highly reliable AI and significantly reduce operational risks.

[1009] The processing flow will be explained below.

[1010] Step 1: Collect data

[1011] The server collects learning data from the company's internal databases and external APIs.

[1012] Action 1.1: Obtain database connection information and establish a connection.

[1013] Action 1.2: Query the data to be collected and retrieve the data. For example, query customer purchase history data.

[1014] Action 1.3: Save the acquired data to temporary storage.

[1015] Step 2: Normalize the data

[1016] The server normalizes the collected data.

[1017] Action 2.1: Standardize data formats, such as date formats and monetary values.

[1018] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[1019] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[1020] Step 3: Discover personal information

[1021] The server detects personal information from the normalized data.

[1022] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[1023] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[1024] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[1025] Step 4: Discover trade secrets

[1026] The server detects trade secrets contained in the data.

[1027] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[1028] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[1029] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[1030] Step 5: Assess data for bias

[1031] The server evaluates the collected data for bias.

[1032] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[1033] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[1034] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[1035] Step 6: Emotion Recognition with the Emotion Engine

[1036] The server recognizes the user's emotions based on the collected data.

[1037] Action 6.1: Analyze audio data. For example, take an audio file and identify emotions such as joy, sadness, or anger from the audio.

[1038] Activity 6.2: Analyze text data, for example customer reviews and feedback, using natural language processing techniques to identify sentiment.

[1039] Action 6.3: List the sentiment analysis results and save them in temporary storage.

[1040] Step 7: Generate reports

[1041] The server generates a report based on the above evaluation results.

[1042] Action 7.1: Compile a list of discovered personal information and trade secrets.

[1043] Action 7.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[1044] Action 7.3: Summarize user feedback based on sentiment analysis results.

[1045] Action 7.4: Output the completed report in PDF or HTML format and save it on the server.

[1046] Step 8: Notify users

[1047] The server notifies the user of the generated report.

[1048] Action 8.1: Create a notification to let the user know where the report is saved.

[1049] Action 8.2: Send a notification to the user via email or corporate notification system.

[1050] Action 8.3: Provide an interface for receiving feedback from the user.

[1051] Example 2

[1052] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1053] Conventional data collection and analysis systems have had problems such as difficulty in identifying personal information or trade secrets contained in the collected training data, and insufficient evaluation of data bias. Furthermore, because data is evaluated without taking user emotions into consideration, the accuracy of the final analysis results is likely to be reduced.

[1054] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1055] In this invention, the server includes means for collecting data from databases and external information sources, means for normalizing the collected data, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for analyzing voice data and text data to recognize user emotions, means for generating a report based on the above evaluation results and emotion recognition results, and means for notifying the user of the generated report. This makes it possible to identify potential risks in the data in advance, take appropriate measures, and perform highly accurate analysis that takes user emotions into consideration.

[1056] A "database" is a computer system that systematically stores information and allows it to be efficiently searched, manipulated, and updated.

[1057] An "external source" is a data source that resides outside the enterprise and is typically accessed via an API.

[1058] "Normalization" is the process of standardizing the format of collected data and filling in missing values.

[1059] "Personal information" is information that can identify a specific individual, and includes, for example, name, address, telephone number, and email address.

[1060] "Trade secrets" are important, non-public information about a company's business, including, for example, information about new products under development and details of internal projects.

[1061] "Data bias" refers to imbalances in a dataset based on specific attributes or conditions.

[1062] "Emotion recognition" is the process of analyzing voice and text data to identify a user's emotions (e.g., joy, sadness, anger, etc.).

[1063] A "report" is a document summarizing detected personal information, trade secrets, data bias, and emotion recognition results.

[1064] "Notification" refers to the act of informing the user of the contents of the generated report, and is carried out via email or an in-company notification system.

[1065] This invention relates to a system that automatically performs processes from data collection to analysis and report generation. This system will be described in detail below.

[1066] Data collection

[1067] The server has a means for collecting training data from the company's internal database and external information sources. For example, the server executes SQL queries against the database to obtain purchase history data. It also obtains data from external information sources using HTTP requests and performs authentication using an API key.

[1068] For example, to retrieve purchase history from the database, the server executes the following SQL query:

[1069] SELECT FROM PurchaseHistory WHERE Date >= '2023-01-01'

[1070] To collect data from an external API, send an HTTP request like this:

[1071] GET https: / / api.example.com / reviews?product_id=123

[1072] Data normalization

[1073] The server normalizes the collected data by standardizing the data format and imputing missing values, such as standardizing date formats to "YYYY-MM-DD," standardizing numeric display formats, and replacing NULL values ​​with the mean or median.

[1074] For example, to standardize date formats, do the following:

[1075] Convert 2023 / 01 / 01 to 2023-01-01

[1076] To fill in missing values, for example, if "purchase amount" is NULL, the overall average purchase amount is substituted.

[1077] Personal information and trade secret detection

[1078] The server uses regular expressions and machine learning models to detect personal information in the normalized data. For example, the regular expression to identify email addresses:

[1079] [a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,4}

[1080] Trade secret detection uses specific keyword lists and machine learning models, for example, scanning data for keywords such as "new product" or "project X."

[1081] Assessing data bias

[1082] The server evaluates the collected data for bias. This evaluation process involves aggregating demographic information such as age, gender, and region, and performing statistical tests. For example, chi-square tests and t-tests are used to verify statistically significant bias. This process can use Python's Matplotlib library to generate histograms and scatter plots.

[1083] emotion recognition

[1084] The server analyzes the voice and text data to recognize the user's emotions. This emotion recognition uses a speech recognition engine (such as the Google Cloud Speech-to-Text API) to convert voice data into text, and natural language processing (NLP) technology to analyze text data.

[1085] As a concrete example, we use Waveform Transform and Google Cloud Speech-to-Text API to analyze audio files and identify user emotions.

[1086] Report generation and notification

[1087] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. This report is generated in PDF or HTML format and saved on the server. For example, the Python ReportLab library is used to generate the PDF report.

[1088] The generated report will be sent to the user via email or the company's internal notification system, for example, a link to the location where the generated report is saved will be sent via email:

[1089] http: / / example.com / reports / Report2023-10-01.pdf

[1090] Specific examples

[1091] For example, suppose a company wants to develop a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Then, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the results. Finally, the report is sent to a person in the company who reviews it and takes any necessary measures.

[1092] Prompt Sentence Examples

[1093] For example, here is an example of a prompt sentence being input to a generative AI model:

[1094] I want to create an AI model that uses customer purchase history data to infer user preferences and interests and propose personalized marketing strategies. How can I build a system that collects data, normalizes it, detects personal information and trade secrets, evaluates data bias, recognizes emotions using an emotion engine, generates reports, and notifies users?

[1095] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1096] Step 1: Collect data

[1097] The server collects data from a company's internal database or external information source. The server first connects to the database and uses authentication information to confirm access rights. For example, it executes an SQL query to retrieve purchase history data. The input is the database connection information and SQL query, and the output is a record of purchase history data. When collecting data from an external information source, the server authenticates using an API key and sends an API request. Specifically, the server sends an HTTP request and temporarily stores the data obtained as a response in storage on the server.

[1098] Step 2: Normalize the data

[1099] The server normalizes the collected data. In this process, the data format is standardized and missing values ​​are filled. For example, the date format is standardized to "YYYY-MM-DD" and NULL values ​​are replaced with the average or median. The input is the collected purchase history data, and the output is the data converted into a standardized format. Specifically, the server applies the rules for data format conversion and missing value filling, and temporarily stores the processed data.

[1100] Step 3: Discover personal information

[1101] The server detects personal information from the normalized data. Regular expressions and machine learning models are used for detection. The input is the normalized data, and the output is a list of detected personal information. Specifically, the server uses predefined regular expressions to identify email addresses and phone numbers, and uses machine learning models to identify addresses and names. The detection results are compiled into a list and stored in temporary storage.

[1102] Step 4: Discover trade secrets

[1103] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for a specific product or the names of internal projects. The input is normalized data, and the output is a list of detected trade secrets. Specifically, the server applies a keyword matching algorithm to extract relevant information. The detection results are listed and temporarily stored.

[1104] Step 5: Assess data for bias

[1105] The server evaluates whether the collected data is biased. This evaluation process involves compiling statistical information such as age, gender, and region, and performing statistical tests. The input is normalized data, and the output is a statistical report indicating whether bias is present or absent. Specifically, the server uses Python libraries to generate histograms and scatter plots and perform chi-square tests and t-tests. This allows the distribution of data to be visualized and statistical bias to be evaluated.

[1106] Step 6: Emotion Recognition

[1107] The server analyzes the voice and text data to recognize the user's emotions. Emotion recognition uses a voice recognition engine and natural language processing (NLP) technology. The input is voice and text data, and the output is analyzed emotion data. Specifically, the server converts the voice data into text (using the Google Cloud Speech-to-Text API) and analyzes the text using a sentiment analysis model. The analysis results are stored in a database.

[1108] Step 7: Generate reports

[1109] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. The input is all of these detection and assessment results, and the output is a report in PDF or HTML format. Specifically, the server uses the Python ReportLab library to generate a PDF report and summarize the report contents. The generated report is saved on the server.

[1110] Step 8: Notify users

[1111] The server notifies the user of the generated report. The input is the generated report and the user's contact information, and the output is a confirmation that the user has been notified. Specifically, the server retrieves the user's email address from a database, generates an email, and sends it using an SMTP server. The email includes a link indicating where the report is saved.

[1112] (Application example 2)

[1113] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the headset type terminal 314 will be referred to as a "terminal."

[1114] In autonomous vehicles, in addition to analyzing driving data and evaluating bias, it is necessary to properly analyze the emotional feedback of the user while driving in order to optimize safety and operation management. Conventional systems lack the functionality to properly evaluate and process this emotional data, which has led to problems such as being unable to adequately evaluate bias in driving data and user stress.

[1115] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1116] In this invention, the server includes means for collecting data from databases and external APIs, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for incorporating an emotion engine that recognizes user emotions, means for generating a report based on the evaluation results and emotion recognition results, and means for notifying the user of the generated report. This enables comprehensive analysis of autonomous vehicle driving data and user emotional feedback, enabling optimization of safety and operation management.

[1117] A "database" is an organized collection of data, a system that stores various data and allows it to be easily searched, retrieved, and analyzed.

[1118] An "external API" is a program interface provided for linking with other software or systems, and is a means of obtaining and manipulating data.

[1119] "Normalization" is the process of standardizing the format of collected data and imputing missing values, improving data quality and consistency.

[1120] "Personal information" is information that identifies a specific individual, such as name, address, and telephone number.

[1121] A "trade secret" is non-public information held by a company or organization, such as technical or commercial information that is kept secret in order to maintain a competitive advantage.

[1122] Bias assessment is the process of detecting and analyzing biases and imbalances present in collected data.

[1123] An "emotion engine" is a technology for recognizing and analyzing a user's emotions, and is a system that estimates their emotional state from voice and text data.

[1124] A "report" is a report summarizing the results of analysis and evaluation of collected data, and is provided to users.

[1125] "Notification" is the process of communicating generated reports and important information to users, typically via email or smartphone notification systems.

[1126] This invention relates to a system for comprehensively analyzing driving data of autonomous vehicles and emotional feedback from users to optimize safety and operation management. The system includes a server, data collection means, data normalization means, personal information detection means, trade secret detection means, bias evaluation means, an emotion recognition engine, report generation means, and user notification means.

[1127] Server Configuration

[1128] The server collects various sensor data (e.g., GPS, speed, battery status, etc.) sent from the autonomous vehicle via a database or external API. The collected data is temporarily stored on the server. Data collection methods include scripts using programming languages ​​such as Python and libraries that call external APIs. For example, the requests library can be used to obtain the required data from external APIs.

[1129] Data normalization measures

[1130] Data normalization involves standardizing the format of the collected data and imputing missing values. To do this, we use the pandas library to manipulate data frames and perform the necessary format conversions and imputation of missing values. For example, we standardize date formats and replace NULL values ​​with the average value.

[1131] Personal information detection method

[1132] Personal information detection techniques identify personal information from normalized data using regular expressions and machine learning models. For example, email addresses and phone numbers can be extracted using regular expressions, and names and addresses can be detected using machine learning models (e.g., models using the scikit-learn library).

[1133] Trade secret detection methods

[1134] The trade secret detection method detects trade secrets in data. It uses a keyword list specified by the company and a machine learning model to detect internal project names and technical information. This also uses the pandas library and natural language processing techniques (e.g., the nltk library and spaCy library).

[1135] Bias assessment measures

[1136] The bias assessment tool aggregates statistical information from the collected data to detect abnormal bias. For example, it creates histograms and scatter plots using the matplotlib and seaborn libraries and performs statistical tests.

[1137] Emotion Recognition Engine

[1138] The emotion recognition engine analyzes voice data and text feedback while driving to identify the user's emotions. To analyze voice data, it uses a voice recognition library (e.g., the SpeechRecognition library), and to analyze the emotions of text data, it uses the TextBlob or VADER library.

[1139] Report Generation Method

[1140] The report generation tool generates reports based on detected personal information, trade secrets, data bias, and sentiment analysis results. These reports are generated in PDF and HTML formats, and are visually easy to understand by utilizing the matplotlib and pdfkit libraries.

[1141] User notification method

[1142] The user notification means notifies the user of the generated report via email or in-app notification. Email is sent using the smtplib library, and a push notification library (e.g., Pushy) is used for in-app notifications.

[1143] Specific examples

[1144] For example, if an autonomous vehicle senses stress due to traffic conditions and finds bias in its driving data, the system will detect the bias and stress in the driving data and provide a detailed report and improvement measures to the operations manager.

[1145] Prompt Sentence Examples

[1146] Please explain in natural language the processes of the "Self-Driving Vehicle Emotion Feedback & Data Analysis System": [Self-Driving Vehicle Sensor Data], [Driving Data Normalization], [Personal Information Anonymization], [Data Bias Assessment], [Emotion Recognition], [Report Generation], and [User Notification].

[1147] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1148] Step 1:

[1149] The server collects sensor data from autonomous vehicles. Input data includes GPS data, speed data, battery data, etc., which are obtained from databases and external APIs. Specifically, the server makes API calls, retrieves data using the requests library, and temporarily stores it. The output is the collected raw data.

[1150] Step 2:

[1151] The server normalizes the data collected in step 1. The input data is raw data, which is stored in a data frame using the PANDAS library. Specific operations include standardizing date formats, converting NULL values ​​to averages, and removing duplicate data. The output is normalized data.

[1152] Step 3:

[1153] The server detects personal information from the normalized data. The input data is normalized data, and regular expressions and machine learning models from the scikit-learn library are used. Specifically, it identifies names, addresses, and phone numbers, creates a list, and saves it in temporary storage. The output is a list of detected personal information.

[1154] Step 4:

[1155] The server detects trade secrets contained in the data. The input data is normalized data, and uses a keyword list specified by the company and natural language processing technology (nltk library and spaCy library). Specifically, it detects specific products and project names, creates a list, and saves it in temporary storage. The output is a list of detected trade secrets.

[1156] Step 5:

[1157] The server evaluates the bias of the collected data. The input data is normalized data, and it is visualized using the matplotlib and seaborn libraries and subjected to statistical testing. Specific operations include creating histograms and scatter plots and determining statistical bias. The output is the bias evaluation results.

[1158] Step 6:

[1159] The server incorporates an emotion engine that recognizes the user's emotions. The input data is voice data and text data from the driver's seat, and it uses a voice recognition library, TextBlob, and VADER library. Specific operations include analyzing voice data, identifying emotions, and identifying emotions from text data. The output is the emotion analysis results.

[1160] Step 7:

[1161] The server generates a report based on the detected personal information, trade secrets, data bias evaluation results, and sentiment analysis results. The input data are the evaluation results obtained at each step, and it uses the matplotlib and pdfkit libraries. Specific operations include visualizing the evaluation results and generating reports in PDF and HTML format. The output is the generated report.

[1162] Step 8:

[1163] The server notifies the user of the generated report. The input data is the generated report, and the smtplib library is used to send emails or push notifications. Specific operations include creating a notification message and sending an email or in-app notification. The output indicates that the notification has been completed.

[1164] The specific processing unit 290 transmits the result of the specific processing to the headset type terminal 314. In the headset type terminal 314, the control unit 46A causes the speaker 240 and the display 343 to output the result of the specific processing. The microphone 238 acquires audio indicating a user input regarding the result of the specific processing. The control unit 46A transmits audio data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the audio data.

[1165] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1166] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the headset type terminal 314.

[1167] [Fourth embodiment]

[1168] FIG. 7 shows an example of the configuration of a data processing system 410 according to the fourth embodiment.

[1169] 7, a data processing system 410 includes a data processing device 12 and a robot 414. An example of the data processing device 12 is a server.

[1170] The data processing device 12 includes a computer 22, a database 24, and a communication I / F 26. The computer 22 is an example of a "computer" according to the technology of the present disclosure. The computer 22 includes a processor 28, a RAM 30, and a storage 32. The processor 28, the RAM 30, and the storage 32 are connected to a bus 34. The database 24 and the communication I / F 26 are also connected to the bus 34. The communication I / F 26 is connected to a network 54. Examples of the network 54 include a WAN (Wide Area Network) and / or a LAN (Local Area Network).

[1171] The robot 414 includes a computer 36, a microphone 238, a speaker 240, a camera 42, a communication I / F 44, and a control target 443. The computer 36 includes a processor 46, a RAM 48, and a storage 50. The processor 46, the RAM 48, and the storage 50 are connected to a bus 52. The microphone 238, the speaker 240, the camera 42, and the control target 443 are also connected to the bus 52.

[1172] The microphone 238 receives instructions and the like from the user 20 by receiving voice uttered by the user 20. The microphone 238 captures the voice uttered by the user 20, converts the captured voice into audio data, and outputs it to the processor 46. The speaker 240 outputs audio in accordance with instructions from the processor 46.

[1173] Camera 42 is a small digital camera equipped with an optical system including a lens, aperture, and shutter, and an imaging element such as a CMOS (Complementary Metal-Oxide-Semiconductor) image sensor or a CCD (Charge Coupled Device) image sensor, and captures images of the surroundings of user 20 (for example, an imaging range defined by an angle of view equivalent to the field of vision of a typical healthy person).

[1174] The communication I / F 44 is connected to a network 54. The communication I / Fs 44 and 26 are responsible for the exchange of various information between the processor 46 and the processor 28 via the network 54. The exchange of various information between the processor 46 and the processor 28 using the communication I / Fs 44 and 26 is carried out in a secure state.

[1175] The control object 443 includes a display device, LEDs in the eyes, and motors for driving the arms, hands, and feet. The posture and gestures of the robot 414 are controlled by controlling the motors of the arms, hands, and feet. Some of the emotions of the robot 414 can be expressed by controlling these motors. In addition, the facial expressions of the robot 414 can also be expressed by controlling the light emission state of the LEDs in the eyes of the robot 414.

[1176] Fig. 8 shows an example of the main functions of the data processing device 12 and the robot 414. As shown in Fig. 8, in the data processing device 12, a specific process is performed by the processor 28. A specific process program 56 is stored in the storage 32.

[1177] The specific processing program 56 is an example of a "program" according to the technology of the present disclosure. The processor 28 reads the specific processing program 56 from the storage 32 and executes the read specific processing program 56 on the RAM 30. The specific processing is realized by the processor 28 operating as a specific processing unit 290 in accordance with the specific processing program 56 executed on the RAM 30.

[1178] The storage 32 stores a data generation model 58 and an emotion identification model 59. The data generation model 58 and the emotion identification model 59 are used by the identification processing unit 290.

[1179] In the robot 414, the processor 46 performs the reception output process. A reception output program 60 is stored in the storage 50. The processor 46 reads the reception output program 60 from the storage 50 and executes the read reception output program 60 on the RAM 48. The reception output process is realized by the processor 46 operating as the control unit 46A in accordance with the reception output program 60 executed on the RAM 48.

[1180] Next, a description will be given of the specific processing performed by the specific processing unit 290 of the data processing device 12. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1181] This invention relates to a system for evaluating and examining AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[1182] Overall Overview

[1183] The server collects the necessary training data from the company's internal database or external API, normalizes it, detects personal information and trade secrets from the normalized data, and evaluates the data for bias. A report is generated based on the evaluation results, and the server notifies the user of the report.

[1184] Details of each step

[1185] Data collection

[1186] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[1187] Data normalization

[1188] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[1189] Personal information detection

[1190] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[1191] Trade Secret Detection

[1192] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[1193] Assessing data bias

[1194] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[1195] Generate reports

[1196] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias. This report includes the detected issues and how to address them. For example, the report is generated in PDF or HTML format and the contents are summarized.

[1197] User Notification

[1198] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[1199] Specific examples

[1200] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. A report is generated based on the evaluation results, and the report is notified to a person in charge within the company. The person in charge reviews the report and takes necessary measures, enabling the development of a reliable AI model.

[1201] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[1202] The processing flow will be explained below.

[1203] Step 1: Collect data

[1204] The server collects learning data from the company's internal databases and external APIs.

[1205] Action 1.1: Obtain database connection information and establish a connection.

[1206] Action 1.2: Execute a query on the data to be collected and retrieve the data. For example, query customer purchase history data.

[1207] Action 1.3: Save the acquired data to temporary storage.

[1208] Step 2: Normalize the data

[1209] The server normalizes the collected data.

[1210] Action 2.1: Standardize data formats, such as date formats and monetary values.

[1211] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[1212] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[1213] Step 3: Discover personal information

[1214] The server detects personal information from the normalized data.

[1215] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[1216] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[1217] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[1218] Step 4: Discover trade secrets

[1219] The server detects trade secrets contained in the data.

[1220] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[1221] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[1222] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[1223] Step 5: Assess data for bias

[1224] The server evaluates the collected data for bias.

[1225] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[1226] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[1227] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[1228] Step 6: Generate reports

[1229] The server generates a report based on the above evaluation results.

[1230] Action 6.1: Compile a list of discovered personal information and trade secrets.

[1231] Action 6.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[1232] Action 6.3: Write a report detailing the problem and its solution.

[1233] Action 6.4: Output the completed report in PDF or HTML format and save it on the server.

[1234] Step 7: Notify users

[1235] The server notifies the user of the generated report.

[1236] Action 7.1: Create a notification to let the user know where to save the report.

[1237] Action 7.2: Send a notification to the user via email or corporate notification system.

[1238] Action 7.3: Provide an interface for receiving feedback from the user.

[1239] Example 1

[1240] Next, a description will be given of Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1241] Conventional data collection and evaluation systems often failed to properly normalize the collected data, and it was difficult to detect personal or confidential information. Furthermore, evaluating whether the data was biased and generating reports based on the evaluation results was time-consuming and prone to human error. Furthermore, notifications of generated reports to users were often done manually, which was inefficient. A highly reliable system was needed to solve these problems.

[1242] The specific processing by the specific processing unit 290 of the data processing device 12 in the first embodiment is realized by the following means.

[1243] In this invention, the server includes means for collecting information from a database or external API to obtain information to be collected, means for normalizing the collected information, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized information, means for detecting confidential information contained in the information, means for evaluating whether the collected information is biased, means for generating a report based on the results of the above evaluations, and means for notifying the user of the generated report.This enables efficient data collection, evaluation, reporting, and notification while increasing the accuracy of the data.

[1244] A "data collection target" is a set of information required for analysis or evaluation.

[1245] A "database" is a digital structure that organizes information and allows it to be searched and updated.

[1246] An "external API" is an interface for accessing external systems and services to obtain data.

[1247] "Information" means data or textual content that is collected, processed, and evaluated for a specific purpose or use.

[1248] "Normalization" is the process of standardizing data formats, removing unnecessary data, and filling in missing values.

[1249] A "missing value" is a value that is not recorded or is missing in a dataset.

[1250] "Personal information" means information that can be used to identify a specific individual, such as name, address, or telephone number.

[1251] "Confidential Information" is non-public, sensitive data that is specific to a particular organization or project.

[1252] "Bias" is a state in which data is overly concentrated or absent in a particular direction or attribute.

[1253] A "report" is a documented result of a particular analysis or evaluation, in a format that allows a user to review the contents.

[1254] "User" is a person or organization that receives reports generated by this system and takes action or makes decisions based on them.

[1255] This invention relates to a system for evaluating and examining AI learning data, and is implemented primarily using a server. Below, the program processing of this system is explained in natural language.

[1256] First, the server uses a database or external API to obtain the information to be collected. Specifically, data is collected using an SQL database (e.g., MySQL, PostgreSQL) or a NoSQL database (e.g., MongoDB). For example, a company's server collects customer purchase history data and temporarily stores it in storage (e.g., Amazon S3).

[1257] The server then normalizes the collected data. This process uses data processing libraries such as Apache Spark and Pandas. Data quality is improved by standardizing data formats, imputing missing values, and removing duplicate data. For example, standardizing date formats and replacing NULL values ​​with a mean value are performed.

[1258] Next, the server detects personal information from the normalized data using regular expressions (e.g., Python's re module) or machine learning models (e.g., spaCy, NLTK). For example, personal information such as name, address, phone number, and email address is detected and temporarily stored as a list.

[1259] Furthermore, the server detects confidential information contained in the data. It extracts confidential information using specific keyword lists and machine learning models (e.g., BERT, GPT). For example, project names and product development information are identified and stored as a list.

[1260] The server then assesses the collected data for bias using Matplotlib and statistical methods (e.g., t-tests, chi-squared tests), visually assessing data distribution with histograms and scatter plots, and performing statistical tests to determine bias.

[1261] Based on the results of the evaluation, the server generates a report in PDF (e.g., ReportLab) or HTML (e.g., Jinja2 template engine) format that includes the issues found and how to address them.

[1262] Finally, the server notifies the user of the generated report via email (e.g., SMTP server) or an internal corporate notification system (e.g., Slack, Microsoft Teams). For example, a link indicating where the generated report is saved may be sent via email.

[1263] Specific examples

[1264] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects purchase history data from a database (e.g., MySQL). Next, it normalizes the collected data using Pandas and detects personal or confidential information using regular expressions and spaCy. It then visualizes the data distribution using Matplotlib and performs statistical tests to evaluate data bias. Based on the evaluation results, it generates a report using ReportLab, and the report is sent to the responsible person via email via an SMTP server. The responsible person reviews the report and takes any necessary measures, enabling the development of a reliable AI model.

[1265] Prompt Sentence Examples

[1266] "Create a program to collect customer purchase history data, normalize the data, and detect personal and confidential information. Also, evaluate the data for bias, generate a PDF report, and notify the user of the report via email."

[1267] This system allows companies to identify potential risks in AI training data in advance and take appropriate measures, making it possible to develop highly reliable AI and significantly reduce operational risks.

[1268] The flow of the identification process in the first embodiment will be described with reference to FIG.

[1269] Step 1:

[1270] The server collects information from a database or external API. It takes a query or API request as input and the collected data as output. For example, to get a user's purchasing history, it runs an SQL query to retrieve the information from the database. The server then temporarily stores this data in storage.

[1271] Step 2:

[1272] The server normalizes the collected data. It requires the collected raw data as input and obtains normalized data as output. Specific operations include standardizing data formats, imputing missing values, and removing duplicate data using Pandas. For example, it standardizes date formats and imputes NULL values ​​with the average value.

[1273] Step 3:

[1274] The server detects personal information from the normalized data. Normalized data is required as input, and the detected personal information is listed as output. Specifically, it uses regular expressions and spaCy to identify personal information such as name, address, phone number, and email address. This information is temporarily stored in storage.

[1275] Step 4:

[1276] The server detects confidential information contained in the data. Normalized data is required as input, and the detected confidential information is listed as output. Specifically, it uses BERT and a specific keyword list to extract confidential information such as project names and product development information. This information is also temporarily stored in storage.

[1277] Step 5:

[1278] The server evaluates whether the collected data is biased. Normalized data is required as input, and the output is the bias evaluation result. Specifically, Matplotlib is used to visually evaluate the data distribution using histograms and scatter plots, and statistical tests such as the chi-square test are performed. The evaluation results are compiled into a report.

[1279] Step 6:

[1280] The server generates a report based on the results of the assessment of detected personal information, confidential information, and bias. It takes as input the results of each detection and assessment, and as output the generated report. It uses ReportLab or the Jinja2 template engine to create reports in PDF and HTML format.

[1281] Step 7:

[1282] The server notifies the user of the generated report. The generated report and the user's contact information are required as input, and a notification is sent to the user as output. Specifically, it sends an email via an SMTP server and provides a link to the location where the report is saved. An in-house notification system can also be used as a notification method.

[1283] (Application example 1)

[1284] Next, a description will be given of Application Example 1. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1285] Sensor data and camera video data collected by autonomous vehicles may contain personal information and trade secrets, so systems are needed to properly process this data while ensuring data accuracy and compliance. Furthermore, if the collected data is biased, it can affect the learning results of the AI ​​model, reducing safety and reliability. Especially for autonomous vehicles, data bias can have a direct impact on the vehicle's driving behavior, so it is necessary to evaluate whether or not there is data bias and correct it appropriately.

[1286] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 1 is realized by the following means.

[1287] In this invention, the server includes means for collecting data from a database or an external API to acquire data to be collected, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for collecting and normalizing sensor data and camera video data from an autonomous vehicle and evaluating whether there is bias in the personal information and data, means for generating a report based on the above evaluation results, and means for notifying a user of the generated report. This makes it possible to evaluate data bias and develop and operate a highly reliable AI model while ensuring data accuracy and compliance in autonomous vehicles.

[1288] A "database" is a set of structured data that allows for efficient storage, management, retrieval, and updating of information.

[1289] An "external API" is an interface for communicating with and exchanging data with other systems or services.

[1290] "Normalization" is the process of standardizing the format of data and imputing missing values.

[1291] "Personal information" is information that can identify a specific individual (e.g., name, address, telephone number, email address).

[1292] "Trade secrets" are information that is kept secret in order to maintain a company's competitive edge, and include development information for specific products and internal project names.

[1293] "Data bias" refers to a situation in which collected data is unevenly distributed across certain attributes (e.g., gender, age, region).

[1294] "Sensor data" refers to data collected from various sensors (e.g., GPS, accelerometer, camera, etc.).

[1295] "Camera video data" refers to still images and video data captured by a camera.

[1296] A "report" is a document that summarizes the results of a particular evaluation or analysis.

[1297] "User" means a person or entity that uses the System.

[1298] An "on-board computer" is a computing device installed in a vehicle and used to process data from sensors and cameras.

[1299] To implement the present invention, a system can be constructed in the following manner: The server performs processes for acquiring, normalizing, detecting, evaluating, generating reports, and notifying data to be collected.

[1300] Program processing and use of hardware and software

[1301] 1. Data Collection

[1302] The server uses a database and external API to collect sensor data and camera video data from the autonomous vehicle, allowing it to understand the vehicle's operating status in real time. For example, it uses the requests library to send HTTP requests and retrieve data.

[1303] 2. Data normalization

[1304] The server performs data normalization to standardize the collected data into a consistent format and impute missing values, using libraries such as pandas and sklearn.preprocessing, including standardizing date formats and scaling sensor data.

[1305] 3. Personal information detection

[1306] The server uses regular expressions and machine learning models to detect personal information from the normalized data, specifically using the re library to identify names, addresses, phone numbers, email addresses, etc.

[1307] 4. Trade Secret Discovery

[1308] The data collected by companies may contain trade secrets, which the server detects using a specific list of keywords and machine learning models.

[1309] 5. Assessing data bias

[1310] The server aggregates statistical information such as gender, age, and region to assess whether the collected data is biased, and performs statistical tests. The pandas library is used to generate histograms and scatter plots to visually evaluate the data distribution.

[1311] 6. Report Generation

[1312] Based on the results of the assessment of detected personal information, trade secrets, and data bias, the server generates a report in PDF or HTML format detailing the issues and how to address them.

[1313] 7. Notice to Users

[1314] Once a report is generated, the user is notified via email or the company's internal notification system, for example by sending a link to the location where the report is saved.

[1315] Specific examples

[1316] Below are some specific examples of data collection and analysis in autonomous vehicles.

[1317] 1. Example prompt:

[1318] The API collects sensor data and camera video data from an autonomous vehicle, retrieves the data every 10 seconds, and stores it on the vehicle's internal computer. The API endpoints are provided below (sensor: http: / / example.com / sensor_data, camera: http: / / example.com / camera_data).

[1319]

[1320] Scale the collected data and standardize the date format to 'YYYY-MM-DD HH:MM:SS'.

[1321]

[1322] Use regular expressions to detect and list names, phone numbers, and email addresses in your data.

[1323]

[1324] Use gender and age demographics to assess whether your dataset is biased.

[1325] In this way, the system can support the development and operation of reliable AI models while ensuring a high level of security and compliance for data collected from autonomous vehicles.

[1326] The flow of the specific processing in the application example 1 will be described with reference to FIG.

[1327] Step 1:

[1328] The server collects sensor data and camera video data from autonomous vehicles from databases and external APIs.

[1329] Input: Database or API endpoint.

[1330] Data processing: Data is retrieved using HTTP requests and temporarily stored in the server's storage. Specifically, data is retrieved in real time using the requests library.

[1331] Output: The raw data collected.

[1332] Step 2:

[1333] The server normalizes the collected data, standardizes it into a consistent format, and imputes missing values.

[1334] Input: Raw data collected.

[1335] Data processing: Using pandas and the sklearn.preprocessing library, we standardized date formats, scaled the data, and imputed missing values ​​with the mean and median.

[1336] Specifically, a data frame is created and the data format of each column is aligned.

[1337] Output: Normalized data.

[1338] Step 3:

[1339] The server detects personal information from the normalized data.

[1340] Input: Normalized data.

[1341] Data processing: Use the re library to extract names, addresses, phone numbers, and email addresses using regular expressions.

[1342] Specifically, pattern matching is performed to create a list of personal information.

[1343] Output: A list of personal information found.

[1344] Step 4:

[1345] The server detects trade secrets from the normalized data.

[1346] Input: Normalized data.

[1347] Data processing: Detecting specific information using company-defined keyword lists and machine learning models.

[1348] Specifically, it loads a machine learning model and makes predictions on the data.

[1349] Output: A list of discovered trade secrets.

[1350] Step 5:

[1351] The server evaluates the collected data for bias.

[1352] Input: Normalized data.

[1353] Data processing: Use the pandas library to aggregate statistics and evaluate the distribution of attribute values ​​such as gender and age.

[1354] Specifically, histograms and scatter plots are generated and biases are detected using statistical methods.

[1355] Output: Report of bias assessment.

[1356] Step 6:

[1357] The server generates a report based on the results of the assessment of detected personal information, trade secrets, and data bias.

[1358] Input: List of personal information, list of trade secrets, bias assessment report.

[1359] Data processing: Compile this information into a document in PDF or HTML format.

[1360] Specifically, data is inserted into an automatically generated template to generate a document.

[1361] Output: Compliance report.

[1362] Step 7:

[1363] The server notifies the user of the generated report.

[1364] Input: Compliance report.

[1365] Data processing: Send a link to the report to users via email or internal notification systems.

[1366] Specifically, it uses an SMTP server to send an email and includes a link.

[1367] Output: Notification email to user.

[1368] Furthermore, an emotion engine that estimates the user's emotion may be further combined. That is, the identification processing unit 290 may estimate the user's emotion using the emotion identification model 59, and perform identification processing using the user's emotion.

[1369] This invention combines an emotion engine with a system that evaluates and examines AI learning data. Below, we will create a program for this system and explain its processing in natural language. We will also provide specific examples.

[1370] Overall Overview

[1371] The server collects the necessary learning data from internal company databases and external APIs and normalizes it. It then detects personal information and trade secrets from the normalized data and evaluates the data for bias. It generates a report based on the evaluation results and notifies the user of the report. The server also incorporates an emotion engine that recognizes user emotions based on the collected data and analyzes user feedback and responses.

[1372] Details of each step

[1373] Data collection

[1374] The server collects learning data from the company's database or external API. For example, if a company wants to develop an AI model using customer purchase history, the server first retrieves the purchase history data from the database. The retrieved data is temporarily stored in the server's storage.

[1375] Data normalization

[1376] The server normalizes the collected data. This process standardizes the data format and imputes missing values ​​(for example, standardizing date formats and replacing NULL values ​​with the mean or median). It also removes duplicate data and unnecessary information.

[1377] Personal information detection

[1378] The server detects personal information from the normalized data using regular expressions and machine learning models. For example, personal information such as name, address, phone number, and email address is identified. The detected personal information is compiled into a list and stored in temporary storage.

[1379] Trade Secret Detection

[1380] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for specific products and internal project names. This information is compiled into a list and stored in temporary storage.

[1381] Assessing data bias

[1382] The server evaluates the collected data for bias. In this evaluation process, statistical information such as age, gender, and region is compiled and the presence or absence of bias is determined through statistical tests. For example, the distribution of data is visually evaluated using histograms and scatter plots, and statistical tests are performed.

[1383] Emotion recognition by emotion engine

[1384] The server recognizes the user's emotions based on the collected data. This emotion recognition involves analyzing voice data and text data. For example, the server analyzes voice data to identify the user's emotions (happiness, sadness, anger, etc.). It also analyzes text data using natural language processing (NLP) techniques to identify the user's emotions.

[1385] Generate reports

[1386] The server generates a report based on the results of the user sentiment analysis, as well as the results of the evaluation of detected personal information and trade secrets, and data bias. This report includes detected problems and how to address them, as well as user feedback based on the sentiment analysis. For example, the report is generated in PDF or HTML format and the contents are summarized.

[1387] User Notification

[1388] The server notifies the user of the generated report via email or the company's internal notification system, for example by sending a link to the location where the generated report is saved.

[1389] Specific examples

[1390] Let's take the example of a company developing a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Next, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the content. Finally, this report is notified to a person in the company who reviews it and takes necessary measures, enabling the development of a reliable AI model.

[1391] This system allows companies to identify potential risks in AI learning data in advance and take appropriate measures, and by taking user emotions into consideration, it is possible to obtain more accurate analysis results. This makes it possible to develop highly reliable AI and significantly reduce operational risks.

[1392] The processing flow will be explained below.

[1393] Step 1: Collect data

[1394] The server collects learning data from the company's internal databases and external APIs.

[1395] Action 1.1: Obtain database connection information and establish a connection.

[1396] Action 1.2: Query the data to be collected and retrieve the data. For example, query customer purchase history data.

[1397] Action 1.3: Save the acquired data to temporary storage.

[1398] Step 2: Normalize the data

[1399] The server normalizes the collected data.

[1400] Action 2.1: Standardize data formats, such as date formats and monetary values.

[1401] Action 2.2: Impute missing or outlier values, e.g., impute NULL values ​​with the mean or median.

[1402] Action 2.3: Clean the data to remove unnecessary and duplicate information.

[1403] Step 3: Discover personal information

[1404] The server detects personal information from the normalized data.

[1405] Action 3.1: Apply regular expressions for personal information detection to extract names, addresses, phone numbers, email addresses, etc.

[1406] Action 3.2: Use a machine learning model to identify potentially personally identifiable information.

[1407] Action 3.3: Create a list of detected personal information and store the results in temporary storage.

[1408] Step 4: Discover trade secrets

[1409] The server detects trade secrets contained in the data.

[1410] Action 4.1: Obtain a list of keywords that should be treated as trade secrets. For example, list the development codes of specific products or internal project names.

[1411] Action 4.2: Apply a keyword matching algorithm to detect trade secrets in the data.

[1412] Action 4.3: Create a list of detected trade secrets and store the detection results in temporary storage.

[1413] Step 5: Assess data for bias

[1414] The server evaluates the collected data for bias.

[1415] Action 5.1: Aggregate statistical information from the data, such as age, gender, and region.

[1416] Activity 5.2: Use visualization tools to visually assess the distribution of data, for example, by generating histograms or scatter plots.

[1417] Action 5.3: Perform statistical tests on potentially biased attributes to determine whether bias exists, for example, a chi-squared test.

[1418] Step 6: Emotion Recognition with the Emotion Engine

[1419] The server recognizes the user's emotions based on the collected data.

[1420] Action 6.1: Analyze audio data. For example, take an audio file and identify emotions such as joy, sadness, or anger from the audio.

[1421] Activity 6.2: Analyze text data, for example customer reviews and feedback, using natural language processing techniques to identify sentiment.

[1422] Action 6.3: List the sentiment analysis results and save them in temporary storage.

[1423] Step 7: Generate reports

[1424] The server generates a report based on the above evaluation results.

[1425] Action 7.1: Compile a list of discovered personal information and trade secrets.

[1426] Action 7.2: Generate a report containing statistical information about the bias in the data and the results of the bias test.

[1427] Action 7.3: Summarize user feedback based on sentiment analysis results.

[1428] Action 7.4: Output the completed report in PDF or HTML format and save it on the server.

[1429] Step 8: Notify users

[1430] The server notifies the user of the generated report.

[1431] Action 8.1: Create a notification to let the user know where the report is saved.

[1432] Action 8.2: Send a notification to the user via email or corporate notification system.

[1433] Action 8.3: Provide an interface for receiving feedback from the user.

[1434] Example 2

[1435] Next, a description will be given of Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1436] Conventional data collection and analysis systems have had problems such as difficulty in identifying personal information or trade secrets contained in the collected training data, and insufficient evaluation of data bias. Furthermore, because data is evaluated without taking user emotions into consideration, the accuracy of the final analysis results is likely to be reduced.

[1437] The specific processing by the specific processing unit 290 of the data processing device 12 in the second embodiment is realized by the following means.

[1438] In this invention, the server includes means for collecting data from databases and external information sources, means for normalizing the collected data, standardizing the data format, and complementing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for analyzing voice data and text data to recognize user emotions, means for generating a report based on the above evaluation results and emotion recognition results, and means for notifying the user of the generated report. This makes it possible to identify potential risks in the data in advance, take appropriate measures, and perform highly accurate analysis that takes user emotions into consideration.

[1439] A "database" is a computer system that systematically stores information and allows it to be efficiently searched, manipulated, and updated.

[1440] An "external source" is a data source that resides outside the enterprise and is typically accessed via an API.

[1441] "Normalization" is the process of standardizing the format of collected data and filling in missing values.

[1442] "Personal information" is information that can identify a specific individual, and includes, for example, name, address, telephone number, and email address.

[1443] "Trade secrets" are important, non-public information about a company's business, including, for example, information about new products under development and details of internal projects.

[1444] "Data bias" refers to imbalances in a dataset based on specific attributes or conditions.

[1445] "Emotion recognition" is the process of analyzing voice and text data to identify a user's emotions (e.g., joy, sadness, anger, etc.).

[1446] A "report" is a document summarizing detected personal information, trade secrets, data bias, and emotion recognition results.

[1447] "Notification" refers to the act of informing the user of the contents of the generated report, and is carried out via email or an in-company notification system.

[1448] This invention relates to a system that automatically performs processes from data collection to analysis and report generation. This system will be described in detail below.

[1449] Data collection

[1450] The server has a means for collecting training data from the company's internal database and external information sources. For example, the server executes SQL queries against the database to obtain purchase history data. It also obtains data from external information sources using HTTP requests and performs authentication using an API key.

[1451] For example, to retrieve purchase history from the database, the server executes the following SQL query:

[1452] SELECT FROM PurchaseHistory WHERE Date >= '2023-01-01'

[1453] To collect data from an external API, send an HTTP request like this:

[1454] GET https: / / api.example.com / reviews?product_id=123

[1455] Data normalization

[1456] The server normalizes the collected data by standardizing the data format and imputing missing values, such as standardizing date formats to "YYYY-MM-DD," standardizing numeric display formats, and replacing NULL values ​​with the mean or median.

[1457] For example, to standardize date formats, do the following:

[1458] Convert 2023 / 01 / 01 to 2023-01-01

[1459] To fill in missing values, for example, if "purchase amount" is NULL, the overall average purchase amount is substituted.

[1460] Personal information and trade secret detection

[1461] The server uses regular expressions and machine learning models to detect personal information in the normalized data. For example, the regular expression to identify email addresses:

[1462] [a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\.[a-zA-Z]{2,4}

[1463] Trade secret detection uses specific keyword lists and machine learning models, for example, scanning data for keywords such as "new product" or "project X."

[1464] Assessing data bias

[1465] The server evaluates the collected data for bias. This evaluation process involves aggregating demographic information such as age, gender, and region, and performing statistical tests. For example, chi-square tests and t-tests are used to verify statistically significant bias. This process can use Python's Matplotlib library to generate histograms and scatter plots.

[1466] emotion recognition

[1467] The server analyzes the voice and text data to recognize the user's emotions. This emotion recognition uses a speech recognition engine (such as the Google Cloud Speech-to-Text API) to convert voice data into text, and natural language processing (NLP) technology to analyze text data.

[1468] As a concrete example, we use Waveform Transform and Google Cloud Speech-to-Text API to analyze audio files and identify user emotions.

[1469] Report generation and notification

[1470] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. This report is generated in PDF or HTML format and saved on the server. For example, the Python ReportLab library is used to generate the PDF report.

[1471] The generated report will be sent to the user via email or the company's internal notification system, for example, a link to the location where the generated report is saved will be sent via email:

[1472] http: / / example.com / reports / Report2023-10-01.pdf

[1473] Specific examples

[1474] For example, suppose a company wants to develop a new AI model using customer purchase history. The company's server first collects the purchase history from a database. Then, it normalizes the collected data, detects personal information and trade secrets, and evaluates the data for bias. It then uses an emotion engine to analyze user feedback and creates a detailed report based on the results. Finally, the report is sent to a person in the company who reviews it and takes any necessary measures.

[1475] Prompt Sentence Examples

[1476] For example, here is an example of a prompt sentence being input to a generative AI model:

[1477] I want to create an AI model that uses customer purchase history data to infer user preferences and interests and propose personalized marketing strategies. How can I build a system that collects data, normalizes it, detects personal information and trade secrets, evaluates data bias, recognizes emotions using an emotion engine, generates reports, and notifies users?

[1478] The flow of the identification process in the second embodiment will be described with reference to FIG.

[1479] Step 1: Collect data

[1480] The server collects data from a company's internal database or external information source. The server first connects to the database and uses authentication information to confirm access rights. For example, it executes an SQL query to retrieve purchase history data. The input is the database connection information and SQL query, and the output is a record of purchase history data. When collecting data from an external information source, the server authenticates using an API key and sends an API request. Specifically, the server sends an HTTP request and temporarily stores the data obtained as a response in storage on the server.

[1481] Step 2: Normalize the data

[1482] The server normalizes the collected data. In this process, the data format is standardized and missing values ​​are filled. For example, the date format is standardized to "YYYY-MM-DD" and NULL values ​​are replaced with the average or median. The input is the collected purchase history data, and the output is the data converted into a standardized format. Specifically, the server applies the rules for data format conversion and missing value filling, and temporarily stores the processed data.

[1483] Step 3: Discover personal information

[1484] The server detects personal information from the normalized data. Regular expressions and machine learning models are used for detection. The input is the normalized data, and the output is a list of detected personal information. Specifically, the server uses predefined regular expressions to identify email addresses and phone numbers, and uses machine learning models to identify addresses and names. The detection results are compiled into a list and stored in temporary storage.

[1485] Step 4: Discover trade secrets

[1486] The server detects trade secrets contained in the data. For example, it uses a keyword list specified by the company and a machine learning model to detect development information for a specific product or the names of internal projects. The input is normalized data, and the output is a list of detected trade secrets. Specifically, the server applies a keyword matching algorithm to extract relevant information. The detection results are listed and temporarily stored.

[1487] Step 5: Assess data for bias

[1488] The server evaluates whether the collected data is biased. This evaluation process involves compiling statistical information such as age, gender, and region, and performing statistical tests. The input is normalized data, and the output is a statistical report indicating whether bias is present or absent. Specifically, the server uses Python libraries to generate histograms and scatter plots and perform chi-square tests and t-tests. This allows the distribution of data to be visualized and statistical bias to be evaluated.

[1489] Step 6: Emotion Recognition

[1490] The server analyzes the voice and text data to recognize the user's emotions. Emotion recognition uses a voice recognition engine and natural language processing (NLP) technology. The input is voice and text data, and the output is analyzed emotion data. Specifically, the server converts the voice data into text (using the Google Cloud Speech-to-Text API) and analyzes the text using a sentiment analysis model. The analysis results are stored in a database.

[1491] Step 7: Generate reports

[1492] The server generates a report based on the detected personal information, trade secrets, data bias assessment results, and emotion recognition results. The input is all of these detection and assessment results, and the output is a report in PDF or HTML format. Specifically, the server uses the Python ReportLab library to generate a PDF report and summarize the report contents. The generated report is saved on the server.

[1493] Step 8: Notify users

[1494] The server notifies the user of the generated report. The input is the generated report and the user's contact information, and the output is a confirmation that the user has been notified. Specifically, the server retrieves the user's email address from a database, generates an email, and sends it using an SMTP server. The email includes a link indicating where the report is saved.

[1495] (Application example 2)

[1496] Next, a description will be given of Application Example 2. In the following description, the data processing device 12 will be referred to as a "server" and the robot 414 will be referred to as a "terminal."

[1497] In autonomous vehicles, in addition to analyzing driving data and evaluating bias, it is necessary to properly analyze the emotional feedback of the user while driving in order to optimize safety and operation management. Conventional systems lack the functionality to properly evaluate and process this emotional data, which has led to problems such as being unable to adequately evaluate bias in driving data and user stress.

[1498] The specific processing by the specific processing unit 290 of the data processing device 12 in the application example 2 is realized by the following means.

[1499] In this invention, the server includes means for collecting data from databases and external APIs, means for normalizing the collected data, standardizing the data format, and imputing missing values, means for detecting personal information from the normalized data, means for detecting trade secrets contained in the data, means for evaluating whether the collected data is biased, means for incorporating an emotion engine that recognizes user emotions, means for generating a report based on the evaluation results and emotion recognition results, and means for notifying the user of the generated report. This enables comprehensive analysis of autonomous vehicle driving data and user emotional feedback, enabling optimization of safety and operation management.

[1500] A "database" is an organized collection of data, a system that stores various data and allows it to be easily searched, retrieved, and analyzed.

[1501] An "external API" is a program interface provided for linking with other software or systems, and is a means of obtaining and manipulating data.

[1502] "Normalization" is the process of standardizing the format of collected data and imputing missing values, improving data quality and consistency.

[1503] "Personal information" is information that identifies a specific individual, such as name, address, and telephone number.

[1504] A "trade secret" is non-public information held by a company or organization, such as technical or commercial information that is kept secret in order to maintain a competitive advantage.

[1505] Bias assessment is the process of detecting and analyzing biases and imbalances present in collected data.

[1506] An "emotion engine" is a technology for recognizing and analyzing a user's emotions, and is a system that estimates their emotional state from voice and text data.

[1507] A "report" is a report summarizing the results of analysis and evaluation of collected data, and is provided to users.

[1508] "Notification" is the process of communicating generated reports and important information to users, typically via email or smartphone notification systems.

[1509] This invention relates to a system for comprehensively analyzing driving data of autonomous vehicles and emotional feedback from users to optimize safety and operation management. The system includes a server, data collection means, data normalization means, personal information detection means, trade secret detection means, bias evaluation means, an emotion recognition engine, report generation means, and user notification means.

[1510] Server Configuration

[1511] The server collects various sensor data (e.g., GPS, speed, battery status, etc.) sent from the autonomous vehicle via a database or external API. The collected data is temporarily stored on the server. Data collection methods include scripts using programming languages ​​such as Python and libraries that call external APIs. For example, the requests library can be used to obtain the required data from external APIs.

[1512] Data normalization measures

[1513] Data normalization involves standardizing the format of the collected data and imputing missing values. To do this, we use the pandas library to manipulate data frames and perform the necessary format conversions and imputation of missing values. For example, we standardize date formats and replace NULL values ​​with the average value.

[1514] Personal information detection method

[1515] Personal information detection techniques identify personal information from normalized data using regular expressions and machine learning models. For example, email addresses and phone numbers can be extracted using regular expressions, and names and addresses can be detected using machine learning models (e.g., models using the scikit-learn library).

[1516] Trade secret detection methods

[1517] The trade secret detection method detects trade secrets in data. It uses a keyword list specified by the company and a machine learning model to detect internal project names and technical information. This also uses the pandas library and natural language processing techniques (e.g., the nltk library and spaCy library).

[1518] Bias assessment measures

[1519] The bias assessment tool aggregates statistical information from the collected data to detect abnormal bias. For example, it creates histograms and scatter plots using the matplotlib and seaborn libraries and performs statistical tests.

[1520] Emotion Recognition Engine

[1521] The emotion recognition engine analyzes voice data and text feedback while driving to identify the user's emotions. To analyze voice data, it uses a voice recognition library (e.g., the SpeechRecognition library), and to analyze the emotions of text data, it uses the TextBlob or VADER library.

[1522] Report Generation Method

[1523] The report generation tool generates reports based on detected personal information, trade secrets, data bias, and sentiment analysis results. These reports are generated in PDF and HTML formats, and are visually easy to understand by utilizing the matplotlib and pdfkit libraries.

[1524] User notification method

[1525] The user notification means notifies the user of the generated report via email or in-app notification. Email is sent using the smtplib library, and a push notification library (e.g., Pushy) is used for in-app notifications.

[1526] Specific examples

[1527] For example, if an autonomous vehicle senses stress due to traffic conditions and finds bias in its driving data, the system will detect the bias and stress in the driving data and provide a detailed report and improvement measures to the operations manager.

[1528] Prompt Sentence Examples

[1529] Please explain in natural language the processes of the "Self-Driving Vehicle Emotion Feedback & Data Analysis System": [Self-Driving Vehicle Sensor Data], [Driving Data Normalization], [Personal Information Anonymization], [Data Bias Assessment], [Emotion Recognition], [Report Generation], and [User Notification].

[1530] The flow of the specific processing in the application example 2 will be described with reference to FIG.

[1531] Step 1:

[1532] The server collects sensor data from autonomous vehicles. Input data includes GPS data, speed data, battery data, etc., which are obtained from databases and external APIs. Specifically, the server makes API calls, retrieves data using the requests library, and temporarily stores it. The output is the collected raw data.

[1533] Step 2:

[1534] The server normalizes the data collected in step 1. The input data is raw data, which is stored in a data frame using the PANDAS library. Specific operations include standardizing date formats, converting NULL values ​​to averages, and removing duplicate data. The output is normalized data.

[1535] Step 3:

[1536] The server detects personal information from the normalized data. The input data is normalized data, and regular expressions and machine learning models from the scikit-learn library are used. Specifically, it identifies names, addresses, and phone numbers, creates a list, and saves it in temporary storage. The output is a list of detected personal information.

[1537] Step 4:

[1538] The server detects trade secrets contained in the data. The input data is normalized data, and uses a keyword list specified by the company and natural language processing technology (nltk library and spaCy library). Specifically, it detects specific products and project names, creates a list, and saves it in temporary storage. The output is a list of detected trade secrets.

[1539] Step 5:

[1540] The server evaluates the bias of the collected data. The input data is normalized data, and it is visualized using the matplotlib and seaborn libraries and subjected to statistical testing. Specific operations include creating histograms and scatter plots and determining statistical bias. The output is the bias evaluation results.

[1541] Step 6:

[1542] The server incorporates an emotion engine that recognizes the user's emotions. The input data is voice data and text data from the driver's seat, and it uses a voice recognition library, TextBlob, and VADER library. Specific operations include analyzing voice data, identifying emotions, and identifying emotions from text data. The output is the emotion analysis results.

[1543] Step 7:

[1544] The server generates a report based on the detected personal information, trade secrets, data bias evaluation results, and sentiment analysis results. The input data are the evaluation results obtained at each step, and it uses the matplotlib and pdfkit libraries. Specific operations include visualizing the evaluation results and generating reports in PDF and HTML format. The output is the generated report.

[1545] Step 8:

[1546] The server notifies the user of the generated report. The input data is the generated report, and the smtplib library is used to send emails or push notifications. Specific operations include creating a notification message and sending an email or in-app notification. The output indicates that the notification has been completed.

[1547] The specific processing unit 290 transmits the result of the specific processing to the robot 414. In the robot 414, the control unit 46A causes the speaker 240 and the control target 443 to output the result of the specific processing. The microphone 238 acquires voice indicating a user input regarding the result of the specific processing. The control unit 46A transmits voice data indicating the user input acquired by the microphone 238 to the data processing device 12. In the data processing device 12, the specific processing unit 290 acquires the voice data.

[1548] The data generation model 58 is a so-called generative AI (Artificial Intelligence). An example of the data generation model 58 is ChatGPT (Internet Search<URL: https: / / openai.com / blog / chatgpt> ), Gemini (Internet search <url: https: gemini.google.com ?hl="ja">) and other generation AIs. The data generation model 58 is obtained by performing deep learning on a neural network. A prompt including an instruction is input to the data generation model 58, and inference data such as voice data indicating voice, text data indicating text, and image data indicating an image is also input. The data generation model 58 performs inference on the input inference data in accordance with the instruction indicated by the prompt, and outputs the inference result in a data format such as voice data and text data. Here, inference refers to, for example, analysis, classification, prediction, and / or summarization.

[1549] In the above embodiment, an example was given in which the specific processing is performed by the data processing device 12, but the technology of the present disclosure is not limited to this, and the specific processing may be performed by the robot 414.

[1550] The emotion identification model 59 as an emotion engine may determine the user's emotion according to a specific mapping. Specifically, the emotion identification model 59 may determine the user's emotion according to an emotion map (see FIG. 9), which is a specific mapping. Similarly, the emotion identification model 59 may determine the robot's emotion, and the identification processing unit 290 may perform identification processing using the robot's emotion.

[1551] FIG. 9 is a diagram illustrating an emotion map 400 on which multiple emotions are mapped. In the emotion map 400, emotions are arranged in concentric circles radiating from the center. Emotions closer to the center of the concentric circles are more primitive. Emotions representing states and actions arising from a state of mind are arranged on the outer edges of the concentric circles. The concept of emotion includes both affect and mental states. Emotions generally generated from reactions occurring in the brain are arranged on the left side of the concentric circles. Emotions generally induced by situational judgment are arranged on the right side of the concentric circles. Emotions generally generated from reactions occurring in the brain and induced by situational judgment are arranged on the upper and lower sides of the concentric circles. Furthermore, the emotion of "pleasure" is arranged on the upper side of the concentric circles, and the emotion of "discomfort" is arranged on the lower side. In this way, in the emotion map 400, multiple emotions are mapped based on the structure by which emotions are generated, and emotions that tend to occur simultaneously are mapped close to each other.

[1552] These emotions are distributed in the 3 o'clock direction on emotion map 400, and typically fluctuate between relief and anxiety. In the right half of emotion map 400, situational awareness dominates over internal sensations, resulting in a sense of calm.

[1553] The inside of emotion map 400 represents what is going on in the mind, and the outside of emotion map 400 represents behavior, so the further you go outside emotion map 400, the more visible the emotions become (the more they are expressed in behavior).

[1554] Human emotions are based on various balances, such as posture and blood sugar levels. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. Emotions can also be created for robots, automobiles, and motorcycles, based on various balances, such as posture and remaining battery life. When these balances deviate from the ideal, a state of discomfort is indicated, and when they approach the ideal, a state of pleasure is indicated. An emotion map can be generated, for example, based on Dr. Mitsuyoshi's emotion map (Research on Voice Emotion Recognition and Emotional Brain Physiological Signal Analysis Systems, Tokushima University, Doctoral Dissertation: https: / / ci.nii.ac.jp / naid / 500000375379). The left half of the emotion map lists emotions belonging to the "reaction" domain, where sensation is dominant. The right half of the emotion map lists emotions belonging to the "situation" domain, where situational awareness is dominant.

[1555] The emotion map defines two emotions that promote learning. One is a negative emotion on the situation side, around the middle of "repentance" or "reflection." In other words, this occurs when the robot experiences negative emotions such as "I never want to feel this way again" or "I don't want to be scolded again." The other is a positive emotion on the response side, around "desire." In other words, this occurs when the robot experiences positive feelings such as "I want more" or "I want to know more."

[1556] The emotion identification model 59 inputs user input into a pre-trained neural network, obtains emotion values ​​indicating each emotion shown in the emotion map 400, and determines the user's emotion. This neural network is pre-trained based on multiple pieces of training data that are combinations of user input and emotion values ​​indicating each emotion shown in the emotion map 400. Furthermore, this neural network is trained so that emotions that are located close to each other have similar values, as in the emotion map 900 shown in FIG. 10. FIG. 10 shows an example in which multiple emotions, "relieved," "calm," and "reassuring," have similar emotion values.

[1557] The system according to the present disclosure has been described above mainly with respect to the functions of the data processing device 12, but the system according to the present disclosure is not necessarily implemented on a server. The system according to the present disclosure may be implemented as a general information processing system. The present disclosure may be implemented, for example, as a software program running on a personal computer or an application running on a smartphone, etc. The method according to the present disclosure may be provided to users in the form of SaaS (Software as a Service).

[1558] In the above embodiment, an example was given in which the specific processing is performed by one computer 22, but the technology of the present disclosure is not limited to this, and the specific processing may be distributed and performed by a plurality of computers including the computer 22. For example, the data generation model 58 may be provided in an external device of the data processing device 12, and data may be generated in the external device in accordance with input data.

[1559] In the above embodiment, an example in which the specific processing program 56 is stored in the storage 32 has been described, but the technology of the present disclosure is not limited to this. For example, the specific processing program 56 may be stored in a portable, computer-readable, non-transitory storage medium such as a USB (Universal Serial Bus) memory. The specific processing program 56 stored in the non-transitory storage medium is installed in the computer 22 of the data processing device 12. The processor 28 executes the specific processing in accordance with the specific processing program 56.

[1560] Alternatively, the specific processing program 56 may be stored in a storage device such as a server connected to the data processing device 12 via the network 54, and the specific processing program 56 may be downloaded and installed on the computer 22 in response to a request from the data processing device 12.

[1561] It is not necessary to store all of the specific processing program 56 in a storage device such as a server connected to the data processing device 12 via the network 54, or to store all of the specific processing program 56 in the storage 32; only a portion of the specific processing program 56 may be stored.

[1562] The hardware resource for executing a specific process can be any of the following processors: An example of a processor is a CPU, which is a general-purpose processor that functions as a hardware resource for executing a specific process by executing software, i.e., a program. Another example of a processor is a dedicated electrical circuit, such as an FPGA (Field-Programmable Gate Array), a PLD (Programmable Logic Device), or an ASIC (Application Specific Integrated Circuit), which is a processor with a circuit configuration designed specifically for executing a specific process. Each processor has built-in or connected memory, and each processor uses the memory to execute the specific process.

[1563] The hardware resource that executes the specific processing may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, the hardware resource that executes the specific processing may be a single processor.

[1564] As an example of a system configured with a single processor, first, one processor is configured by combining one or more CPUs and software, and this processor functions as a hardware resource that executes a specific process. Second, there is a system that uses a processor that realizes the functions of an entire system including multiple hardware resources that execute a specific process on a single IC chip, as typified by SoC (System-on-a-chip). In this way, a specific process is realized using one or more of the above-mentioned various processors as hardware resources.

[1565] Furthermore, the hardware structure of these various processors can be, more specifically, an electric circuit that combines circuit elements such as semiconductor devices. The specific processing described above is merely an example. Therefore, it goes without saying that unnecessary steps may be deleted, new steps may be added, or the processing order may be rearranged, without departing from the spirit of the invention.

[1566] The above-described description and illustrations are a detailed explanation of the parts related to the technology of the present disclosure and are merely an example of the technology of the present disclosure. For example, the above description of the configuration, functions, actions, and effects is an explanation of an example of the configuration, functions, actions, and effects of the parts related to the technology of the present disclosure. Therefore, it goes without saying that unnecessary parts may be deleted, new elements may be added, or replacements may be made to the above-described description and illustrations within the scope of the gist of the technology of the present disclosure. Furthermore, to avoid confusion and facilitate understanding of the parts related to the technology of the present disclosure, the above-described description and illustrations omit explanations of common technical knowledge that do not require particular explanation to enable the implementation of the technology of the present disclosure.

[1567] All publications, patent applications, and technical standards mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent application, or technical standard was specifically and individually indicated to be incorporated by reference.

[1568] The following is further disclosed regarding the above embodiment.

[1569] (Claim 1)

[1570] A means of collecting data from databases and external APIs to obtain the data to be collected;

[1571] A means to normalize the collected data, standardize the data format, and impute missing values;

[1572] A means for detecting personal information from the normalized data;

[1573] a means for detecting trade secrets contained in the data;

[1574] A means of assessing whether the collected data is biased; and

[1575] A means for generating a report based on the above evaluation results;

[1576] The system includes a means for notifying a user of a generated report.

[1577] (Claim 2)

[1578] 2. The system according to claim 1, further comprising a means for standardizing the date format and the format of monetary amounts as a normalization means for standardizing the data format and filling in missing values.

[1579] (Claim 3)

[1580] The system of claim 1, further comprising a means for using regular expressions or machine learning models as a means for detecting personal information from normalized data.

[1581] "Example 1"

[1582] (Claim 1)

[1583] A means of collecting information from databases and external APIs to obtain the information to be collected;

[1584] A means to normalize the collected information, standardize the data format, and impute missing values;

[1585] A means for detecting personal information from the normalized information;

[1586] means for detecting confidential information contained within the information;

[1587] A means of assessing whether the collected information is biased; and

[1588] A means for generating a report based on the above evaluation results;

[1589] The system includes a means for notifying users of generated reports.

[1590] (Claim 2)

[1591] 2. The system according to claim 1, further comprising a means for standardizing the date format and the format of numeric display as a normalization means for standardizing the data format and filling in missing values.

[1592] (Claim 3)

[1593] The system according to claim 1, further comprising a means for using regular expressions or machine learning models as a means for detecting personal information from normalized information.

[1594] "Application Example 1"

[1595] (Claim 1)

[1596] A means of collecting data from databases and external APIs to obtain the data to be collected;

[1597] A means to normalize the collected data, standardize the data format, and impute missing values;

[1598] A means for detecting personal information from the normalized data;

[1599] a means for detecting trade secrets contained in the data;

[1600] A means of assessing whether the collected data is biased; and

[1601] A means to collect and normalize sensor data and camera video data from autonomous vehicles and evaluate the presence or absence of personal information and data bias;

[1602] A means for generating a report based on the above evaluation results;

[1603] The system includes a means for notifying a user of a generated report.

[1604] (Claim 2)

[1605] 2. The system according to claim 1, further comprising a means for standardizing the date format and the format of monetary amounts as a normalization means for standardizing the data format and filling in missing values.

[1606] (Claim 3)

[1607] The system of claim 1, further comprising a means for using regular expressions or machine learning models as a means for detecting personal information from normalized data.

[1608] "Example 2: Combining Emotion Engines"

[1609] (Claim 1)

[1610] The means of collecting data from databases and external sources to obtain the data to be collected;

[1611] A means to normalize the collected data, standardize the data format, and impute missing values;

[1612] A means for detecting personal information from the normalized data;

[1613] a means for detecting trade secrets contained in the data;

[1614] A means of assessing whether the collected data is biased; and

[1615] A means for recognizing user emotions by analyzing voice data and text data;

[1616] A means for generating a report based on the evaluation results and emotion recognition results;

[1617] The system includes a means for notifying the generated report.

[1618] (Claim 2)

[1619] 2. The system according to claim 1, further comprising a means for standardizing the date format and the format of numeric display as a normalization means for standardizing the data format and filling in missing values.

[1620] (Claim 3)

[1621] The system of claim 1, further comprising a means for using regular expressions or machine learning models as a means for detecting personal information from normalized data.

[1622] "Application example 2 when combining emotion engines"

[1623] (Claim 1)

[1624] A means of collecting data from databases and external APIs,

[1625] A means to normalize the collected data, standardize the data format, and impute missing values;

[1626] A means for detecting personal information from the normalized data;

[1627] a means for detecting trade secrets contained in the data;

[1628] A means of assessing whether the collected data is biased; and

[1629] a means for incorporating an emotion engine that recognizes user emotions;

[1630] A means for generating a report based on the evaluation results and emotion recognition results;

[1631] The system includes a means for notifying a user of a generated report.

[1632] (Claim 2)

[1633] 2. The system according to claim 1, further comprising a means for standardizing the date format and the format of monetary amounts as a normalization means for standardizing the data format and filling in missing values.

[1634] (Claim 3)

[1635] The system of claim 1, further comprising a means for using regular expressions or machine learning models as a means for detecting personal information from normalized data. [Explanation of symbols]

[1636] 10, 210, 310, 410 Data Processing Systems 12 Data Processing Device 14 Smart Devices 214 Smart Glasses 314 Headset-type terminal 414 Robot< / url:> < / url:> < / url:> < / url:>

Claims

1. A means of collecting data from databases and external APIs to obtain the data to be collected; A means to normalize the collected data, standardize the data format, and impute missing values; A means for detecting personal information from the normalized data; a means for detecting trade secrets contained in the data; A means of assessing whether the collected data is biased; and A means for generating a report based on the above evaluation results; The system includes a means for notifying a user of a generated report.

2. 2. The system according to claim 1, further comprising a means for standardizing the date format and the format of monetary amounts as normalization means for standardizing the data format and filling in missing values.

3. The system according to claim 1 , further comprising a means for using regular expressions or machine learning models as a means for detecting personal information from normalized data.

Images