system

Abstract

Provide a system. An observation information acquisition means for acquiring observation information consisting of environmental information including structures, moving objects, and brightness conditions in a number of spatial regions; a hearing information acquisition means for acquiring hearing information consisting of opinion information provided by local stakeholders; an instruction generating means for generating an instruction for predicting regional environmental characteristics after a plurality of future periods based on the observation information and the hearing information; a prediction information acquisition means for inputting the instruction sentence generated by the instruction sentence generation means into a generative AI model and acquiring prediction information indicating the regional environmental characteristics after the multiple periods from the generative AI model; An index information output means for generating and outputting index information applicable to resource allocation policy or business plan formulation based on the forecast information. A system including:

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 a description and related instruction sentence regarding 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] JP 2022-180282 A Summary of the Invention [Problem to be solved by the invention]

[0004] Traditionally, future predictions for the local environment have relied on limited publicly available statistical information and a few indicators, which means that only superficial factors are considered, and there is a problem that the actual situation in the region is not accurately reflected. In addition, non-quantitative information from local residents and stakeholders has not been fully utilized, making it difficult to provide accurate guidelines for mid- to long-term investment plans and strategy planning. As a result, there have been issues that have led to errors in resource allocation and business development, and have hindered the realization of regional revitalization and improved quality of life for residents. [Means for solving the problem]

[0005] In this invention, the server includes: [an observation information acquisition means for acquiring observation information consisting of environmental information including structures, moving objects, and luminance states in multiple spatial regions]; [an interview information acquisition means for acquiring interview information consisting of opinion information provided by local stakeholders]; [an instruction statement generation means and a prediction information acquisition means for generating an instruction statement for predicting local environmental characteristics after multiple future periods based on the observation information and the interview information and inputting the instruction statement to a generative AI model]; and [an index information output means for outputting index information applicable to resource allocation policy or business plan formulation based on the prediction information]. This makes it possible to accurately grasp the actual state of the local environment and derive optimal strategies and investment plans from a medium- to long-term perspective.

[0006] "Observation information" refers to environmental information including structures, moving objects and brightness conditions in multiple spatial regions, and is information that objectively and externally indicates the actual state of the target area. "Hearing information" refers to opinion information provided by local stakeholders, which includes subjective and experiential knowledge about the local environment. An "instruction sentence" is a sentence expressing a command or request that is input to a generative AI model based on the observational information and the interview information, and includes conditions and indicators for predicting future regional environmental characteristics from the observational information and the interview information. A "generative AI model" is a trained information processing means that generates information for predicting future regional environmental characteristics in response to input instructions, and is a model that operates using artificial intelligence technology including natural language processing. "Predictive information" refers to information indicating future regional environmental characteristics that is output by the generative AI model in response to the instruction text, and provides guidelines that will be useful in formulating resource allocation policies and business plans. "Index information" is information that indicates quantitative or qualitative criteria to be used in determining resource allocation policies or business plan formulation based on forecast information. [Brief description of the drawings]

[0007] [Figure 1] FIG. 2 is a sequence diagram showing the flow of processing of the system in the present embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, an example of an embodiment of a system according to the technology of the present disclosure will be described with reference to the accompanying drawings.

[0009] <System configuration>

[0010] An example of an embodiment of the present invention will be described below. Note that the configuration described below is merely an example and can be modified without departing from the spirit of the present invention.

[0011] In this embodiment, the present invention is embodied as a system that predicts future characteristics of a local environment based on observed information and interviewed information, and generates index information for formulating resource allocation policies or business plans. This system is realized by a configuration including at least a "server" and a "terminal."

[0012] The server is an information processing device, and a general computer device can be used as a program execution environment. For example, the server is preferably configured to include multiple processing devices (CPU, GPU, etc.), a main storage device, an auxiliary storage device, and a network interface. The server may have a database for receiving and storing observation information (environmental information including structures, moving objects, and brightness conditions in multiple spatial regions) and hearing information (opinion information obtained from local stakeholders) provided from the terminal. The server has a function of generating instruction sentences based on the observation information and hearing information, and inputting the instruction sentences into a generative AI model to obtain prediction information regarding future local environmental characteristics. The server also generates index information applicable to resource allocation policies or business plan formulation based on the prediction information, and transmits it to the terminal described later. The server includes a software module for executing this series of processes. For example, a natural language processing means, an image analysis means, a time series analysis means, and an interface module for utilizing the generative AI model are implemented to analyze the observation information and hearing information.

[0013] The terminal is an information processing device capable of communicating with the server via a network. The terminal can be equipped with a plurality of acquisition means, such as a camera, a microphone, a brightness sensor, a mobile sensor, and a temperature and humidity sensor, as observation information acquisition means. The terminal executes a program (e.g., an image analysis library, a voice recognition engine, a script for shaping sensor values) for preprocessing the information acquired from these acquisition means. The terminal acquires the observation information, and also acquires the hearing information by directly contacting the local stakeholders and recording and converting their remarks into text. A general voice recognition engine can be used for the text conversion. The terminal transmits the observation information and the hearing information thus acquired to the server by a protocol such as HTTP. The terminal also receives the index information transmitted from the server, visualizes it on a map display screen or a dashboard-type UI, and presents it to the user. The terminal can be implemented, for example, as a mobile computing device, a tablet terminal, a mobile information terminal, or a stationary information processing device with a display attached thereto.

[0014] The communication means for connecting the server and the terminals may be the Internet or a local network environment, regardless of whether it is wired or wireless. The server stores the observational information and the hearing information sent from the terminals in a memory area and performs preprocessing on the information (missing value completion, keyword extraction from text, object detection in images, time axis analysis, etc.). The server then uses the preprocessing results to generate prompts (instructions) and inputs them into a generative AI model to obtain forecast information that predicts regional environmental characteristics 2 to 3 years into the future. The server further generates indicator information from this forecast information, such as investment priorities by region, the need for infrastructure development, and an estimate of the investment payback period. When the terminal receives this indicator information, it visualizes it on a user interface, allowing users (regional government officials, private businesses, infrastructure project stakeholders, etc.) to develop effective resource allocation policies and business plans based on the information.

[0015] With this configuration, the system of this embodiment can comprehensively utilize observational information and interview information to accurately predict future regional environmental characteristics, and can provide users with information to make decisions based on that prediction.

[0016] <Example of system configuration>

[0017] The server and terminals configured in this embodiment will be described in more detail below.

[0018] (Server embodiment) The server is a core information processing device for integrating a wide variety of information related to the local environment and predicting future local environmental characteristics. The server can be configured as a general information processing device, and includes, for example, multiple processing units (CPU, GPU), a main memory (DRAM), an auxiliary memory (HDD, SSD), and a network interface. An operating system (UNIX (registered trademark) type or other suitable operating system) runs on the server, and multiple software modules are executed on it.

[0019] The server includes an analysis means for processing the observation information and the hearing information received from the terminal. The observation information is environmental information including structures (buildings, facilities, etc.), moving objects (vehicles, bicycles, pedestrians, etc.) and luminance conditions (nighttime illuminance, horizontal light source density, etc.) in the spatial domain of the target area. The server has a database for storing this observation information, and the database is associated with time series data, location information, category information, etc. The hearing information is information including subjective opinions and experience knowledge obtained from local stakeholders, and the server stores this in the database as text data. The server processes this observation information and hearing information in an integrated manner, and uses natural language processing means, image analysis means, and time series analysis means as preprocessing for making the desired future prediction.

[0020] As a natural language processing means, the server extracts keywords, performs sentiment analysis, and classifies themes from text data (interviewed information). For example, if a local resident says something like "the number of stores open at night is increasing," the server extracts keywords such as "nighttime business," "increase," and "store" from the text and treats them as elements suggesting changes in the local environment. As an image analysis means, the server uses image and video data contained in the observation information to analyze the density of people flowing through the city, the deterioration of buildings, and the distribution of vehicle types. Time series analysis means are used to analyze fluctuations in this data over time, and can extract changes such as "there has been an increasing trend in young people going out late at night over the past six months." These preprocessing results are used as material for generating instruction sentences that the server inputs into the generative AI model.

[0021] Based on these analysis results, the server generates instructions for predicting regional environmental characteristics two to three years into the future. The instructions are input information for the generative AI model, and include the target region, time period, and factors to be considered (behavior patterns of young people, infrastructure development status, utilization rate of commercial facilities, etc.). The server sends these instructions to the generative AI model, and receives predicted information in response. The generative AI model is an advanced trained model based on natural language processing, and makes future predictions based on the input instructions.

[0022] The server further reprocesses the obtained forecast information to generate index information applicable to resource allocation policy or business plan formulation. Examples of this index information include infrastructure investment priorities in a specific region, the effectiveness of commercial facility development, location plans for new services aimed at young people, etc. The server transmits this index information to the terminal, enabling it to be visualized and presented on the terminal.

[0023] (Embodiment of terminal) The terminal is an information processing device that acquires observational and hearing information in the local environment and transmits it to the server. The terminal may be configured to be able to connect to multiple external sensors such as a camera, a microphone, an illuminance sensor, a temperature and humidity sensor, and a mobile object detection sensor. For example, the terminal can be a portable small computer device that can be placed in the observation area, whether indoors or outdoors. The camera acquires images or videos of the target location or streetscape, and the microphone collects sounds and voices on the street. The illuminance sensor quantifies the brightness at night, and the mobile object detection sensor can measure the number of pedestrians and vehicles passing by.

[0024] The terminal also has an interface for conducting interviews with local residents and stakeholders. For example, the voice of the person being interviewed can be recorded using a microphone connected to the terminal, and then converted to text using voice recognition software. The textual information is then basically reformatted (converted into a data format such as JSON) inside the terminal, and then sent to a server. The terminal communicates with the server via a communication line (wireless LAN, mobile communication network, wired line, etc.), and uploads the observation information and interview information to the server periodically or upon a specified trigger (e.g., the passage of a certain period of time, detection of a specific event, etc.).

[0025] The terminal may have a function to visualize the indicator information received from the server on a user interface. For example, a regional map may be displayed on the display screen of the terminal, and the investment priority and the degree of infrastructure development necessity for each region may be color-coded and mapped. In addition, an interface that makes it easy for users to plan strategies based on the obtained indicator information can be provided using text-based reports and graph displays. The terminal may be configured as a tablet, mobile device, laptop computer, or stationary workstation, and can be selected according to the user's operability and the size of the display screen.

[0026] As described above, by organically linking the terminal and the server, in this embodiment, future predictions can be made by effectively utilizing observational information and interview information that reflects the actual local environment, and the results can be provided to users as information that is useful for formulating resource allocation policies and business plans.

[0027] <System Operation>

[0028] The server executes the program of the system using Python 3.8 running on a Linux environment. The server stores the observation information and the hearing information in a structured format using a MySQL database management system, and functions as a web server that receives HTTP requests from terminals using Nginx. The server analyzes the observation information and the hearing information using software such as spaCy for natural language processing, statsmodels for time series analysis, and OpenCV for image analysis, and generates prompts for the generative AI model using the analysis results. The server also obtains future predictions using generative AI models such as OpenAI GPT-4 API, and derives indicator information that contributes to resource allocation policies or business plan formulation from the prediction information.

[0029] The terminal executes Python scripts on a small computer such as Raspberry Pi (registered trademark) or a tablet terminal, and acquires observation information using a Logitech (registered trademark) camera, microphone, light sensor, temperature and humidity sensor, etc. The terminal converts the contents of the hearing into text using a voice recognition engine such as the Whisper (registered trademark) model, and transmits the observation information and the hearing information to the server via an HTTP request. The terminal receives the index information returned from the server in JSON format, and presents it to the user via a map display tool such as Leaflet (registered trademark) or Mapbox (registered trademark) or a user interface using HTML / CSS / JavaScript (registered trademark).

[0030] Users can visually refer to indicator information such as investment priorities by region, infrastructure development needs, and payback periods on their device screens. By referring to this information, users can formulate resource allocation policies and investment plans.

[0031] (Example of a prompt) "Below is observational data from the past two years in a certain regional area (traffic volume growth rate, changes in the nighttime brightness index, and increased late-night activity among young people) as well as opinions from local residents and business owners that 'areas where young people gather at night are becoming more active.' Based on this data, make a comprehensive prediction and summarize the possibility of commercial area expansion, infrastructure development needs, and changes in the residential environment in this area over the next two to three years."

[0032] <Example of the system's operation (Figure 1)>

[0033] Step 1: The terminal uses a camera, microphone, light sensor, and temperature and humidity sensor to obtain observation information (video data, audio data, brightness values, weather values) and listening information (audio opinions from local stakeholders). The terminal receives signals from the above sensor devices (camera images, recorded audio, sensor values) as input, processes them using Python scripts to perform image analysis using OpenCV, audio input and output using pyaudio (registered trademark), and environmental value acquisition through I2C communication, and generates formatted observation information and listening information (including text-converted audio information) as output. This output data is converted into JSON format for subsequent processing and sent to the server.

[0034] Step 2: The server receives the observational and listening information (input) sent from the terminal via HTTP requests using Nginx as a front end, and stores it in a MySQL database. The server receives the input data in JSON format, and performs data processing and calculations such as formatting the time series data using pandas (registered trademark), extracting text using a natural language processing tool (spaCy), and associating it with information analyzed using OpenCV, and obtains a reformated analyzable data set as output.

[0035] Step 3: The server uses time series analysis (StatsModels (registered trademark)) on the formatted data set (input) to analyze trends from the past to the present and extracts fluctuation trends and features. The server performs data calculations on this input data, such as calculating traffic volume fluctuation rates, frequency analysis of times when young people are present, and time averaging of brightness values, and obtains as output an extended data set that includes a set of features required for future predictions (traffic growth rates, nighttime brightness transition indexes, etc.).

[0036] Step 4: The server uses the extended data set (input) to automatically generate prompts for the generative AI model. Here, the server performs data calculations that comprehensively consider keywords extracted from the textualized interview information (such as "increase in nighttime business hours" and "young people using late-night hours") and changing trends in the city obtained from image analysis, and creates, as output, instruction sentences (prompt sentences) that can be input to the generative AI model. An example of this prompt sentence is, "The following is observational data from the past two years in a certain local area (traffic volume growth rate, nighttime brightness index fluctuations, and increase in young people's late-night activities) and opinions obtained from local residents and businesses. Based on this data, comprehensively predict and summarize the possibility of commercial area expansion, infrastructure development needs, and changes in the residential environment in this area in two to three years' time."

[0037] Step 5: The server sends the above prompt text (input) to the generative AI model and obtains future prediction information (output) as a response from the generative AI model. The server accesses the generative AI model via the OpenAI GPT-4 API and receives text information indicating the regional environmental characteristics two to three years into the future (for example, "Area A has a high priority for commercial investment, while Area B requires infrastructure improvements in the medium term") as a response based on the input prompt text.

[0038] Step 6: The server analyzes the received future forecast information (input) and extracts and generates indicator information useful for resource allocation policies or business plans. The server performs text mining and rule-based analysis on this input information, and obtains indicator information in JSON format as output, including specific investment planning indicators (for example, "Area A: High priority for commercial investment (5-year payback period)" and "Area B: Mid-term need for infrastructure reinforcement (7-year payback period)").

[0039] Step 7: The terminal visualizes the index information (input) received from the server using map display tools such as HTML / CSS / JavaScript or Leaflet, and presents it to the user. The terminal analyzes the input index information, processes the data by displaying color-coded areas on a map, or displaying summaries in graph or text format, and generates a screen display that the user can directly check as output. The user can then make decisions about the allocation of regional resources and investment strategies based on the results of this display.

Claims

1. an observation information acquisition means for acquiring observation information consisting of environmental information including structures, moving objects, and luminance conditions in a number of spatial regions; a hearing information acquisition means for acquiring hearing information consisting of opinion information provided by local stakeholders; an instruction generating means for generating an instruction for predicting a regional environmental characteristic after a plurality of future periods based on the observation information and the hearing information; a prediction information acquisition means for inputting the instruction sentence generated by the instruction sentence generation means into a generative AI model and acquiring prediction information indicating the regional environmental characteristics after the multiple periods from the generative AI model; An index information output means for generating and outputting index information applicable to resource allocation policy or business plan formulation based on the forecast information. A system including:

2. The system described in claim 1, wherein the instruction sentence generation means further comprises means for analyzing the observation information and the interview information by period to extract regional environmental change trends, and dynamically changing the prompt sentence to be input to the generative AI model based on the extraction results of the regional environmental change trends.

3. The system of claim 1, wherein the prediction information acquisition means comprises means for organizing the prediction information obtained from the generative AI model by regional environmental characteristics, and continuously generating a plurality of prompt sentences for formulating resource allocation policies or business plans in accordance with the results of organizing the prediction information.

Images