Method and system for automatically generating customized format report based on AI

Abstract

The invention discloses a method and system for automatically generating a customized format report based on AI, and relates to the technical field of artificial intelligence, and the method comprises the steps: collecting user data, and carrying out the preprocessing of the collected user data; based on the preprocessed user data, constructing an emotion analysis model, and identifying a user emotion state; according to an analysis result of the sentiment analysis model, calculating to obtain a report content unit priority score, and performing content unit priority judgment; generating a customized format report based on the content unit priority judgment result and a user customization demand; and performing encryption processing on the user data. By combining the sentiment analysis model, calculating the priority score of the content unit, and based on the sentiment state and the content demand of the user, each part of the report is scored and sorted, so that the content which most accords with the user intention is ensured to be preferentially displayed, the problems that the report content is disordered and does not accord with the user demand are avoided, and the user experience is improved. Therefore, the structuralization and logicality of the report are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, and particularly relates to a method and system for automatically generating a customized format report based on AI. BACKGROUND

[0002] In recent years, artificial intelligence AI technology has been widely applied in various industries, especially in the aspects of automatic report generation and natural language processing NLP, the progress of AI is particularly remarkable. Traditional report generation methods mainly rely on manual sorting and writing, which not only consumes a lot of time and effort, but also cannot quickly respond to the personalized needs of users. In order to improve efficiency, AI is gradually applied to various links of report generation, from preliminary data collection, text processing to automatic generation of format. In the prior art, a natural language generation NLG system can realize partial automatic report generation through template matching and preset rules.

[0003] The existing report generation technology has obvious limitations in many aspects. First, most of these technologies lack recognition and analysis of user emotional state, which makes it difficult for the generated report to accurately reflect the real intention and needs of the user. As an advanced AI technology, sentiment analysis can identify the emotional state of the user in different scenarios, thereby realizing personalized content adjustment and priority setting during report generation. However, the application of sentiment analysis in the prior art is relatively preliminary, usually limited to simple sentiment classification, and cannot combine sentiment weight to optimize the judgment of content priority. At the same time, the calculation of content priority in the prior art is mostly a static model, lacking a mechanism for dynamic adjustment according to user needs, resulting in a lack of flexibility and customization in the structure and format of the generated report. SUMMARY

[0004] In view of the problems existing in the prior art of automatically generating a customized format report based on AI, the present application is proposed.

[0005] Therefore, the problem to be solved by the present application is that the existing method for generating a customized format report lacks recognition and analysis of the user's emotional state, which makes it difficult for the generated report to accurately reflect the real intention and needs of the user.

[0006] To solve the above technical problems, the present application provides the following technical scheme: a method for automatically generating a customized format report based on AI, comprising: collecting user data, preprocessing the collected user data; based on the preprocessed user data, constructing a sentiment analysis model to identify the user's emotional state; according to the analysis result of the sentiment analysis model, calculating the priority score of the report content unit to determine the priority of the content unit; based on the priority determination result of the content unit and the user's customized needs, generating a customized format report; and encrypting the user data.

[0007] As a preferred scheme of the method for automatically generating a customized format report based on AI, wherein: the collection of user data includes, Collecting user customization requirements for reports, including content preferences, format requirements, and structure preferences, and extracting user voice data and text data from user customization requirements; The voice data refers to the collection of user voice signal data using a high-sensitivity microphone; The text data refers to the collection of user written reports, notes, and historical reports.

[0008] As a preferred scheme of the method for automatically generating a customized format report based on AI, wherein: the pre-processing of the collected user data includes, Performing voice data processing and text data processing; The voice signal data processing refers to using a low-pass filter and a mean filter to remove background noise to obtain filtered voice signals; Frame and windowing of voice signals include, Set frame length L frame and frame shift L shift , frame the continuous voice signal according to the set frame length L frame and frame shift L shift , to obtain a sequence of voice signals of multiple time frames; Using a Hamming window function, multiply each frame of voice signal sequence with a Hamming window function to generate windowed voice signals; Based on the windowed voice signal, perform short-time Fourier transform STFT to obtain the time-frequency spectrum Q(ω, t) of the voice signal; Using the Mel Frequency Cepstral Coefficient MFCC extraction method to extract the features of the voice signal to generate a voice feature vector MFCC t , the voice feature vector MFCC t includes pitch, volume, and speech rate; Using automatic speech recognition technology ASR to convert voice data into corresponding text data; The text data processing refers to using the Stanford CoreNLP tool to divide the text into different sentences and paragraphs, each independent sentence and paragraph being a text unit, using a pre-processed BERT model to vectorize each text unit to generate a text embedding vector X; Concatenate the voice feature vector MFCC t and the text embedding vector X to obtain a multi-modal feature vector X f .

[0009] As a preferred scheme of the method for automatically generating a customized format report based on AI, wherein: the constructing of the sentiment analysis model based on the preprocessed user data, the identification of the user emotional state comprises, The sentiment analysis model is constructed using a long short-term memory network (LSTM), including an input layer, an embedding layer, an LSTM layer, a full connection layer, and an output layer. Collect historical user data for feature extraction and generate historical multi-modal feature vectors X fl , and input the historical multi-modal feature vectors X fl into the sentiment analysis model as training data and update the parameters using an Adam optimizer. Set the sentiment categories to positive, neutral, and negative, and input the multi-modal feature vectors X f generated by the user in real time into the trained sentiment analysis model, output the probability of each sentiment category using a Softmax function, and generate a sentiment score vector Ei.

[0010] As a preferred scheme of the method for automatically generating a customized format report based on AI, wherein: the constructing of the sentiment analysis model based on the preprocessed user data, the identification of the user emotional state comprises, Based on the positive sentiment probability P positive , the sentiment weight W E is calculated. The context correlation C(x i , y j ) between the text units is calculated using the cosine similarity formula. The information entropy H(x i ) of the text unit x i is calculated using the information entropy calculation formula. The time weight W T (t) is calculated. Based on the sentiment weight W E , the context correlation C(x i , y j ), the time weight W T (t), and the information entropy H(x i ), the priority score P i of each text unit is calculated. i The priority score P i is calculated as follows. The text unit priority score P i of each content unit is calculated comprehensively based on the text unit priority score P i , and the content unit priority score Pl is obtained. The content unit priority score Pl is compared with the content unit priority score threshold Pl' for judgment. If Pl>Pl', it is determined that the content unit is in a high priority level; If Pl≤Pl', it is determined that the content unit is in a low priority level.

[0011] As a preferred scheme of the method for automatically generating a customized format report based on AI, wherein: the generating of the customized format report based on the content unit priority determination result and the user customization requirement comprises, According to the user's customized format requirement, set different priority content unit format rules, including, Title style, paragraph style and mark important content, combine content unit and format rule, generate formatted content; According to the text unit priority score Pi, the screened content is laid out to generate a customized format report.

[0012] As a preferred scheme of the method for automatically generating a customized format report based on AI, wherein: the encryption processing of the user data comprises, The user registers an account and sets a password, generates a random salt value according to the set password, and splices the generated salt value with the user set password; The spliced password is used as the input of the hash algorithm SHA-256, the hash value of the password is calculated by using the hash algorithm SHA-256, and the generated hash value and the used salt value are stored in the database; When the user logs in, the password is verified, the stored salt value is read from the database, and the user input password and the salt value are spliced, the hash value is recalculated, and the newly calculated hash value is compared with the hash value in the database; If the hash values are consistent, the password verification is passed; If the hash values are inconsistent, automatically record the verification time and the error reason and send them to the user in real time.

[0013] Another object of the present application is to provide a system for automatically generating a customized format report based on AI, which comprises, A data acquisition module for acquiring user data; A preprocessing module for preprocessing the acquired user data; An emotion analysis module for constructing an emotion analysis model and analyzing user emotions; A data analysis module for calculating report content priority score and generating a customized format report; An encryption module for encrypting user data.

[0014] A computer device comprises a memory and a processor; the memory stores a computer program, and the processor implements the steps of the method for automatically generating a customized format report based on AI when executing the computer program.

[0015] A computer readable storage medium stores a computer program, and the computer program implements the steps of the method for automatically generating a customized format report based on AI when executed by a processor.

[0016] The present application has the beneficial effects that: the present application calculates the priority score of the content unit by combining the sentiment analysis model, scores and sorts each part of the report based on the emotional state and content demand of the user, to ensure that the content most consistent with the user's intention is preferentially displayed, avoiding the problem of disordered report content and inconsistency with user demand, thereby improving the structured and logical nature of the report. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 The flowchart of the method for automatically generating a customized format report based on AI.

[0019] Figure 2 The flowchart of generating a customized format report.

[0020] Figure 3 The structure diagram of the system for automatically generating a customized format report based on AI. DETAILED DESCRIPTION

[0021] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings.

[0022] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the scope of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0023] Second, the "one embodiment" or "an embodiment" referred to herein can include a particular feature, structure, or characteristic. The various embodiments appearing at different places in the specification are not necessarily all cumulative or mutually exclusive of each other.

[0024] Embodiment 1: Reference Figure 1 and Figure 2 As a first embodiment of the present application, the embodiment provides a method for automatically generating a customized format report based on AI, the method for automatically generating a customized format report based on AI comprising, S1, collecting user data, and pre-processing the collected user data; Further, collecting user data includes, collecting user customization requirements for reports, including content preferences, format requirements, and structure preferences, and extracting user voice data and text data from user customization requirements; The voice data refers to collecting user voice signal data using a high-sensitivity microphone; The text data refers to collecting user written reports, notes, and historical reports.

[0025] By collecting user voice signal data using a high-sensitivity microphone, detailed capture of user voice emotion and tone change is achieved. The role of voice data is not only to provide more dimensional information for emotion analysis, but also to assist in identifying emotional fluctuations when users express, thereby making report generation more in line with user emotional state and current needs. In addition, as part of natural interaction, voice data reduces the user's operational burden, making the entire customization process more natural and smooth. Furthermore, by collecting user written reports, notes, and historical reports, a large amount of text information is provided, which can provide important reference for the user's long-term preferences for the system, further optimizing the generation logic of the report.

[0026] Further, pre-processing the collected user data includes, performing voice data processing and text data processing; The voice signal data processing refers to using a low-pass filter and a mean filter to remove background noise to obtain filtered voice signals; Frame and windowing of voice signals include, Setting frame length L frame to 2Ams, frame shift L shift to Ams, and performing frame operation on the continuous voice signals according to the set frame length L frame and frame shift L shift to obtain a sequence of voice signals of multiple time frames; Use the Hamming window function to multiply each frame of the speech signal sequence by the Hamming window function to generate a windowed speech signal; Based on the windowed speech signal, a short-time Fourier transform (STFT) is performed to obtain the time-frequency spectrum Q(ω, t) of the speech signal, where ω is the frequency of the speech signal and t is the time. Use Mel-frequency cepstral coefficient MFCC extraction method to extract the characteristics of speech signal and generate speech feature vector MFCC t , the speech feature vector MFCC t Including pitch, volume and speaking speed, the MFCC extraction formula is: ; Among them, MFCC k is the kth MFCC coefficient, H m (ω) is the mth filter of the Mel filter bank, and DCT is discrete cosine transform; Use automatic speech recognition technology ASR to convert voice data into corresponding text data; The text data processing refers to using the Stanford CoreNLP tool to divide the text into different sentences and paragraphs, each independent sentence and paragraph is a text unit, and using the pre-processed BERT model to vectorize each text unit to generate a text embedding vector X; The speech feature vector MFCC t Concatenate the vector X with the text embedding vector to obtain the multimodal feature vector X f .

[0027] By using low-pass and mean filters to remove background noise, the clarity and accuracy of the speech signal are ensured, ensuring that subsequent speech feature extraction can be performed in a clean speech signal environment, thereby improving the accuracy of speech features. Furthermore, framing and windowing operations decompose the continuous speech signal into multiple short time frames, and a Hamming window function is used to reduce the impact of spectral leakage. This step makes the speech features within each time frame more representative, facilitating frequency domain analysis and feature extraction of the speech signal. ASR technology is used to convert the speech signal into text data, achieving efficient conversion between speech and text. The use of ASR technology not only reduces manual input time but also improves the practicality and efficiency of the system. During the text data processing stage, the Stanford CoreNLP tool is used to segment the text into sentences and paragraphs, making the text processing highly structured. After dividing the text into independent text units, each unit can be further vectorized to ensure that different types of information can be effectively represented in vector form, providing a foundation for subsequent sentiment analysis and priority determination.

[0028] S2, based on the pre-processed user data, constructing a sentiment analysis model to identify the user emotional state; Further, based on the pre-processed user data, constructing a sentiment analysis model to identify the user emotional state includes, Using a long short-term memory network (LSTM) to construct a sentiment analysis model, including an input layer, an embedding layer, an LSTM layer, a fully connected layer, and an output layer; Collecting historical user data for feature extraction and generating historical multi-modal feature vectors X fl , inputting the historical multi-modal feature vectors X fl into the sentiment analysis model as training data and updating the parameters using an Adam optimizer; Setting the sentiment categories to positive, neutral, and negative, and inputting the real-time collected user-generated multi-modal feature vectors X f into the trained sentiment analysis model to output the probability of each sentiment category using a Softmax function: ; Where P(y=k| h T ) is the probability of the input data belonging to category k, z k is the score of category k, K is the total number of categories, h T is the output of the fully connected layer of the sentiment analysis model, and z j is the score of category j, and y is the category; Combining the probability of each sentiment category to generate a sentiment score vector Ei: ; Where P positive , P neutral , and P negative represent the probabilities of positive, neutral, and negative sentiment, respectively, and P positive + P neutral + P negative = 1.

[0029] By constructing a sentiment analysis model, the user's emotional state is directly incorporated into the parameters of content generation, and by identifying the user's emotional state, the user's emotional needs can be more accurately reflected, so that report generation no longer simply relies on the logic and priority of the content, but can dynamically combine the user's current emotional state to adjust the content arrangement and key information presentation of the report. The effect not only improves the user relevance of the report, but also makes the content and form of the report more in line with the user's psychological expectations, ultimately enhancing the user's satisfaction and acceptance of the generated report. By analyzing the user's emotional changes, the system can adjust the report content generation strategy in real time and at stages, enhancing the real-time and flexibility of report generation, allowing the system to maintain high adaptability in complex and changing user emotional scenarios.

[0030] S3、According to the analysis result of the sentiment analysis model, the report content unit priority score is calculated, and the content unit priority is determined; Further, according to the analysis result of the sentiment analysis model, the report content unit priority score is calculated, and the content unit priority is determined, which includes, Based on the positive sentiment probability P positive Calculate the sentiment weight W E : ; Where, ∆P(t) is the change rate of positive sentiment probability, t is the generation time of the current text unit, T is the time period of generating the report, and ∆P(t)=P positive (t)-P positive (t-1), only according to the positive sentiment probability P positive cannot reflect the change and dynamic fluctuation of user sentiment, the change rate of positive sentiment probability ∆P(t) makes us can capture the change of user sentiment with time or content input, the user's emotion is not linear change, it is affected by many factors, including context, time and different input content, through the nonlinear function sin( P positive (t)dt) mapping sentiment score, simulating the complex fluctuation of human emotion, using the adaptive function η(∆P) of the change rate of positive sentiment probability ∆P(t) can dynamically adjust the weight according to the sentiment change, so that the content unit with large sentiment fluctuation can automatically get higher priority ranking, η(∆P)=e α∣∆P(t)∣ , α is the adjustment coefficient, used to control the sensitivity of sentiment change, when the sentiment change is small, η(ΔP) grows slowly, which shows linear adjustment, when the sentiment change is significant, the exponential function will grow rapidly, and the influence of fluctuation on weight will be enhanced; The cosine similarity formula is used to calculate the context correlation C(x i , y j ) between text units, and the context correlation C(x i , y j ) calculation formula is: ; Where, x i and x j are the vector representation of text units i and j respectively, the value range of cosine similarity is [−1,1], and the closer the value of context correlation is to 1, the higher the correlation is; The information entropy H(x i ) of text unit x i is calculated using the information entropy calculation formula, and the formula is: ; Among them, H(x i ) is the text unit x i The information entropy, p(w j ) is the jth word w j In text cell x i The word frequency in , n is the total number of words in the text unit; Calculate the time weight W T (t): ; Where u is the integral variable, ranging from 0 to t, that is, from the start time to the current time, and ω is the angular frequency of the periodic time T, ω = 2π / T. By using the sine function sin(ωu), the periodic changes in time influence can be captured. The linear attenuation of existing technologies cannot reflect this periodic change, and the simple exponential decay formula cannot step the nonlinear growth and decay of time weight. The use of the logarithmic function ln(1+u) and the inverse term of time 1 / 1+u can simultaneously characterize the nonlinear growth and decay of time weight, which has better adaptability to scenarios with long-term cumulative effects and initial rapid decay. The derivative term du / dt captures the dynamic changes of time, allowing the time weight to be adaptively adjusted without manually setting the attenuation parameters and weight factors. The existing time weight model can only handle a single time decay mode. The optimized time weight calculation formula can simultaneously consider multiple time factors by introducing multiple functions, such as rapid changes in the short term, cumulative effects in the long term, and periodic fluctuations. Based on the sentiment weight W E , context relevance C(x i ,y j ), time weight W T (t) and information entropy H(x i ), calculate the priority score P of each text unit i , priority score P i The calculation formula is as follows: ; Where λ is the balance coefficient, which is used to balance the impact of information entropy on priority scoring. λ is set by experts based on industry experience and related field characteristics. According to the text unit priority score P i Give each content unit a text unit priority score P i Perform comprehensive calculations to obtain the content unit priority score Pl; Based on historical experience and relevant industry standards, experts set a content unit priority score threshold Pl' and compare it with the content unit priority score Pl; If Pl > Pl', it is determined that the content unit is of high priority, and detailed display is performed at the front of the report; If Pl < Pl', it is determined that the content unit is of low priority, and simplified display is performed in the report.

[0031] By scoring the priority of each content unit, the system can optimize the logical structure of the report, and the generation of the report is no longer a simple accumulation of information, but an orderly arrangement of content through the scoring algorithm, ensuring that important and relevant content is displayed to the user first. This optimization not only improves the efficiency of information transmission, but also helps users quickly find key information and reduces the interference of redundant content. For example, key decision-making information and content highly relevant to the user's emotional state are placed in a prominent position in the report, allowing users to obtain valuable information in a short period of time. In addition, the priority scoring mechanism can be dynamically adjusted according to the user's real-time needs and emotional changes. When the user's emotional state changes, the system can update the priority of the content unit in real time, thereby dynamically updating the content of the report. Compared with static report generation methods, this method is more flexible and adaptable, and can adjust the content structure according to the user's emotional changes, enhancing the personalized recommendation effect of the report. For example, when the user has a higher demand for a certain type of information, the system can automatically increase the priority of related content based on the score. This dynamic response mechanism greatly improves the intelligence level of report generation, and the ordering of report content units is based on the user's emotions and the relevance of the content, thereby forming a hierarchical report that allows users to naturally transition from important information to secondary information when reading the report. The entire report structure is clear and logically rigorous, greatly improving the user's reading experience. This hierarchical structure allows the report to remain well-organized even when the amount of information is large, reducing the risk of information overload.

[0032] S4, generating a customized format report based on the content unit priority determination result and the user's customized needs; Further, generating a customized format report based on the content unit priority determination result and the user's customized needs includes, According to the user's customized format needs, setting different priority content unit format rules, including, Title style, paragraph style, and marking important content, combining content units and format rules to generate formatted content; According to the text unit priority score Pi, the formatted content is laid out to generate a customized format report.

[0033] Through sentiment analysis and priority determination, the system can identify which content units are most important to the user, thereby prioritizing the display of these contents, which not only ensures the relevance of the report content, but also reduces the user's time and effort in browsing unnecessary information, making the report better convey key information. Different users may have significantly different needs for report formats, for example, in the education field, the report may need to be more academic and structured, while in the marketing field, the report format may need to be more creative and visually impactful. Therefore, by combining the user's customized needs, the system can generate personalized reports according to the user's specific preferences, such as layout, font, color, and content layout, which not only makes the content display flexible and easy, but also meets the user's specific operation habits and aesthetic needs, improving the user's interactive experience with the report. At the same time, as the user's needs and priorities change, the report format can be dynamically adjusted in real time, and the user does not need to set the format every time, which significantly improves the operation convenience and reduces the need for user intervention, greatly improving the user's satisfaction and the system's usability, fully meeting the diverse and personalized needs of users in different industries and scenarios.

[0034] S5, encrypting the user data; Further, the encryption of the user data includes, The user registers an account and sets a password, generates a random salt value according to the set password, and concatenates the generated salt value with the user's set password; Use the concatenated password as the input of the SHA-256 hash algorithm, calculate the hash value of the password using the SHA-256 hash algorithm, and store the generated hash value and the used salt value in the database; When the user logs in, verify the password, read the stored salt value from the database, concatenate the user's input password with the salt value, recalculate the hash value, and compare the newly calculated hash value with the hash value in the database; If the hash values are consistent, the password verification is passed; If the hash values are not consistent, automatically record the verification time and error reason and send them to the user in real time.

[0035] By selecting a hash algorithm for password encryption, by converting user data into a fixed length hash value, even a small change in data will cause the generated hash value to change significantly, therefore, the system can quickly detect whether the data has been tampered with by comparing the hash values of the original data and the received data when transmitting and storing data, so that the integrity of the data can be effectively verified without encrypting the actual data content, enhancing the anti-tamper ability of the system, ensuring the consistency and reliability of user data throughout the processing process, effectively preventing these data from being leaked or used for illegal purposes without user authorization, improving the privacy protection level of the system, and enhancing the trust of users in the system.

[0036] Embodiment 2: Refer to Figure 3 As a second embodiment of the present application, this embodiment is different from the previous embodiment, and provides a system for automatically generating a customized format report based on AI, comprising, A data acquisition module for acquiring user data; A preprocessing module for preprocessing the acquired user data; An emotion analysis module for constructing an emotion analysis model to analyze user emotions; A data analysis module for calculating the priority score of the report content and generating a customized format report; An encryption module for encrypting user data.

[0037] If the functions are realized in the form of software function units and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application or the parts that essentially contribute to the prior art or parts of the technical solutions can be embodied in the form of software products, which are stored in a storage medium and include instructions to make a computer device (which can be a personal computer, server, or network device) execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various program code storage media.

[0038] The logic and / or steps represented in the flow diagrams or otherwise described herein, for example, can be considered as a sequence of executable instructions, and can be embodied in any computer-readable medium for use by or in connection with an instruction execution system, apparatus, or device, such as a computer-based system, processor-containing system, or other system that can fetch the instructions from the instruction execution system, apparatus, or device and execute the instructions. In the context of this specification, a "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device. The computer-readable medium can be, for example but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, device, or propagation medium.

[0039] More specific examples (a non-exhaustive list) of the computer-readable medium include the following: an electrical connection (electronic) having one or more wires, a portable computer diskette (magnetic), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can also be paper or another suitable medium upon which the program is printed, as the program can be electronically captured, for example via an optical scanner, then compiled, interpreted, or otherwise processed, and stored in a computer memory in a form that can be later executed by a computer.

[0040] It should be understood that aspects of the application can be implemented in hardware, software, firmware or combinations thereof. In the above embodiments, various steps or methods can be implemented in software or firmware that is stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, any of the following technologies, known in the art, or combinations thereof, can be used: a discrete logic circuit having logic gates for implementing logic functions upon data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array (PGA), a field programmable gate array (FPGA), or the like.

[0041] It should be noted that the above-mentioned embodiments are merely used to illustrate the technical solutions of the present application but not limit the present application, and although the present application is described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalent replaced without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for automatically generating customized format reports based on AI, characterized by: include, Collect user data and pre-process the collected user data; Based on the pre-processed user data, a sentiment analysis model is constructed to identify the user's emotional state; According to the analysis results of the sentiment analysis model, the priority score of the report content unit is calculated and the priority of the content unit is determined; Generate customized format reports based on content unit priority determination results and user customization requirements; Encrypt user data.

2. The method for automatically generating a customized report based on AI according to claim 1, wherein: The collection of user data includes: Collect user customized requirements for reports, including content preferences, format requirements, and structure preferences, and extract user voice data and text data based on user customized requirements; The voice data refers to the user's voice signal data collected using a high-sensitivity microphone; The text data refers to the collection of users' written reports, notes and historical reports.

3. The method for automatically generating a customized report based on AI according to claim 2, wherein: The pre-processing of the collected user data includes: Perform voice data processing and text data processing; The speech signal data processing refers to removing background noise using a low-pass filter and a mean filter to obtain a filtered speech signal; Framing and windowing of speech signals include: Set the frame length L frame and frame shift L shift , the continuous speech signal is transmitted according to the set frame length L frame and frame shift L shift Performing a framing operation to obtain a speech signal sequence of multiple time frames; Use the Hamming window function to multiply each frame of the speech signal sequence by the Hamming window function to generate a windowed speech signal; Based on the windowed speech signal, short-time Fourier transform (STFT) is performed to obtain the time-frequency spectrum Q(ω, t) of the speech signal; Use Mel-frequency cepstral coefficient MFCC extraction method to extract the characteristics of speech signal and generate speech feature vector MFCC t , the speech feature vector MFCC t including pitch, volume, and speed of speech; Use automatic speech recognition technology ASR to convert voice data into corresponding text data; The text data processing refers to using the Stanford CoreNLP tool to divide the text into different sentences and paragraphs, each independent sentence and paragraph is a text unit, and using the pre-processed BERT model to vectorize each text unit to generate a text embedding vector X; The speech feature vector MFCC t Concatenate the vector X with the text embedding vector to obtain the multimodal feature vector X f .

4. The method for automatically generating a customized report based on AI according to claim 3, wherein: The method of constructing a sentiment analysis model based on pre-processed user data and identifying the user's emotional state includes: Use the long short-term memory network (LSTM) to build a sentiment analysis model, including the input layer, embedding layer, LSTM layer, fully connected layer, and output layer; Collect historical user data for feature extraction and generate historical multimodal feature vector X fl , the historical multimodal feature vector X fl Input the sentiment analysis model as training data and use the Adam optimizer to update the parameters; Set the emotion categories to positive, neutral and negative, and generate the multimodal feature vector X based on real-time user acquisition. f Input the trained sentiment analysis model and generate the sentiment score vector Ei by outputting the probability of each sentiment category using the Softmax function.

5. The method for automatically generating a customized report based on AI according to claim 4, characterized in that: The analysis results of the sentiment analysis model are used to calculate the priority score of the report content unit and determine the priority of the content unit, including: Based on the positive sentiment probability P positive Calculate the sentiment weight W E ; The cosine similarity formula is used to calculate the contextual correlation between text units C(x i ,y j ); Use the information entropy calculation formula to calculate the text unit x i The information entropy H(x i ); Calculate the time weight W T (t); Based on the sentiment weight W E , context relevance C(x i ,y j ), time weight W T (t) and information entropy H(x i ), calculate the priority score P of each text unit i ; According to the text unit priority score P i Give each content unit a text unit priority score P i Perform comprehensive calculation to obtain the content unit priority score Pl; Establish a content unit priority score threshold Pl' and compare it with the content unit priority score Pl; If Pl>Pl', it is determined that the content unit is at a high priority; If Pl≤Pl', it is determined that the content unit is at a low priority.

6. The method for automatically generating a customized report based on AI according to claim 5, characterized in that: The generating of a customized format report based on the content unit priority determination result and the user's customized requirements includes: According to the user's customized format requirements, set content unit format rules of different priorities, including: Heading styles, paragraph styles, and marking important content combine content units and formatting rules to generate formatted content; According to the text unit priority score Pi, the filtered content is laid out and a customized format report is generated.

7. The method for automatically generating a customized report based on AI according to claim 6, characterized in that: The encryption process of user data includes: The user registers an account and sets a password. A random salt value is generated based on the set password, and the generated salt value is concatenated with the user's set password. Use the concatenated password as the input of the hash algorithm SHA-256, calculate the hash value of the password using the hash algorithm SHA-256, and store the generated hash value and the used salt value in the database; When the user logs in, the password is verified, the stored salt value is read from the database, the password entered by the user is concatenated with the salt value, the hash value is recalculated, and the newly calculated hash value is compared with the hash value in the database; If the hash values ​​are consistent, the password verification is successful; If the hash values ​​are inconsistent, the verification time and error reason will be automatically recorded and sent to the user in real time.

8. A system for automatically generating customized format reports based on AI, based on the method for automatically generating customized format reports based on AI according to any one of claims 1 to 7, characterized in that: include, Data collection module, used to collect user data; A preprocessing module, used to preprocess the collected user data; Sentiment analysis module, used to build sentiment analysis models and analyze user emotions; Data analysis module, used to calculate the priority score of report content and generate customized format reports; The encryption module is used to encrypt user data.

9. A computer device comprising: A memory and a processor; the memory stores a computer program, characterized in that: when the processor executes the computer program, the steps of the method for automatically generating a customized format report based on AI according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for automatically generating a customized format report based on AI according to any one of claims 1 to 7 are implemented.

Images