Information generation method and device, electronic equipment, storage medium and program product

Abstract

The invention discloses an information generation method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of computers, and the method comprises the steps that a message database is constructed, and the message database is used for storing messages; and behavior data of the historical processing message of the target user is acquired, and the user demand of the target user is determined according to the behavior data. And generating personalized message information corresponding to the target user based on the user demand. In this way, a user can be helped to quickly extract the core key points in the message and obtain important information in the message, it is ensured that the user can efficiently master and use the core information in the message, the efficiency and quality of information absorption by the user are greatly improved, then the user can efficiently process the message, and the energy and time of reading the message are reduced.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an information generation method, apparatus, electronic device, storage medium, and program product. Background Technology

[0002] As society continues to develop, users need to read a large number of reports (or messages) every day, which consumes a lot of time and energy. However, existing automated text processing technologies cannot efficiently process messages, causing users to spend a lot of time and energy reading them. Summary of the Invention

[0003] This application provides an information generation method, apparatus, electronic device, storage medium, and program product that can efficiently process messages.

[0004] In a first aspect, embodiments of this application provide an information generation method, including: Construct a message database, which is used to store messages; Obtain historical behavioral data of the target user in processing the aforementioned messages; The user needs of the target user are determined based on the behavioral data. Based on the user's needs, a personalized message is generated for the target user.

[0005] Optionally, the personalized message information includes at least one of message annotation and message summary, wherein the message annotation is generated based on the message summary.

[0006] Optionally, generating personalized message information corresponding to the target user based on the user's needs includes: Generate at least one candidate personalized message based on the user's requirements; The at least one candidate personalized message is sorted based on the word-limited reward SBWR algorithm, and the candidate personalized message located at the preset position is taken as the personalized message corresponding to the target user.

[0007] Optionally, the behavioral data includes at least one of the following: the target user's historical comments on the message, the target user's historical messages viewed, the target user's historical message keywords retrieved, the target user's historical click data, and the target user's historical activity trajectory; The step of determining the user needs of the target user based on the behavioral data includes: determining the user needs corresponding to the behavioral data through a pre-trained intent recognition model; Before determining the user demand corresponding to the behavioral data through a pre-trained intent recognition model, the method further includes: The behavioral data is input into the intent recognition model to obtain the first output result; Add noise to the first output result to obtain the second output result; Based on the first output result, the second output result, and the preset output result, the cross-entropy value and the relative entropy value corresponding to the intent recognition model are determined; the cross-entropy value is used to determine the difference between the first output result and the preset output result, and the relative entropy value is used to determine the difference between the second output result and the preset output result; The intent recognition model is trained based on the cross-entropy value and the relative entropy value until the cross-entropy value is less than a second preset threshold and / or the relative entropy value is less than a third preset threshold, thereby obtaining the pre-trained intent recognition model.

[0008] Optionally, the method further includes: If the target message includes numerical data corresponding to time data, the numerical data is predicted to determine the numerical data of the target time, and the numerical data of the target time is displayed. The target message is the message corresponding to the personalized message information.

[0009] Optionally, the numerical data for displaying the target time includes: The personalized message information displays numerical data of the target time.

[0010] Optionally, the method further includes: The personalized message information is converted into an audio file and then played.

[0011] Secondly, embodiments of this application also provide an information generation apparatus, including: The construction module is used to build a message database, which is used to store messages; The acquisition module is used to acquire historical behavior data of the target user in processing the message; The determination module is used to determine the user needs of the target user based on the behavioral data; The generation module is used to generate personalized message information corresponding to the target user based on the user's requirements.

[0012] Thirdly, embodiments of this application provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor, wherein when the program is executed by the processor, it implements the steps of the information generation method as described in any of the first aspects.

[0013] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the information generation method as described in any of the first aspects.

[0014] Fifthly, embodiments of this application also provide a computer program product, which is stored in a storage medium and executed by at least one processor to implement the steps of the information generation method as described in any of the first aspects.

[0015] In this embodiment, the electronic device can acquire historical message processing behavior data of the target user, determine the user's needs based on the behavior data, and then generate personalized message information corresponding to the target user based on the user's needs. This helps users quickly extract the core points of the message, obtain important information from the message, and ensure that users can efficiently grasp and use the core information in the message, greatly improving the efficiency and quality of information absorption, thereby enabling users to process messages efficiently and reduce the effort and time spent reading messages. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of an information generation method according to an embodiment of this application; Figure 2 This is a flowchart of a message storage method provided in an embodiment of this application; Figure 3 This is a schematic diagram of an information generation device according to an embodiment of this application.

[0018] Figure 4 This is a schematic diagram of an electronic device according to an embodiment of this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] With the increasing number of reports (or messages) such as work reports, government reports, industry reports, and departmental reports, people often need to spend a lot of time reading, studying, and summarizing these messages. The sheer volume of messages makes it difficult for users to quickly find and understand the key information, significantly impacting policy understanding, work progress, industry direction, and the grasp and implementation of key departmental tasks.

[0021] However, existing technologies cannot efficiently process messages to help users quickly find and understand key information and reduce reading time. Specifically, existing automated text processing tools may not be able to efficiently classify, sort, and extract key information from large volumes of messages. Furthermore, different users have different focuses and needs regarding messages, but existing technologies cannot achieve efficient personalized information extraction and summarization to specifically help users extract key information and improve message reading efficiency.

[0022] This application provides an information generation method, apparatus, electronic device, storage medium, and program product. The embodiments of this application will be described in detail below with reference to the accompanying drawings and specific embodiments and application scenarios.

[0023] Please see Figure 1 , Figure 1 This is a flowchart of an information generation method provided in an embodiment of this application. The method includes the following steps: Step 101: Construct a message database, which is used to store messages.

[0024] In some embodiments, the electronic device acquires imported messages to construct a message database. The message import method can be either automatic import of messages according to a format provided by the electronic device, or manual import of messages in batches by the user on the electronic device.

[0025] In some embodiments, electronic devices can manage collected messages in batches. For example, electronic devices can approve messages, store messages, add messages, edit messages, view messages, and delete messages.

[0026] The following is a detailed explanation of message approval and message entry: Electronic devices can send erroneous messages to the corresponding departments. After the departments correct the erroneous messages, they can upload the corrected messages back to the electronic devices. After the approval process confirms that the messages are correct, the messages are then entered into the message database.

[0027] The following is a specific workflow illustrating the message approval and message storage process, such as... Figure 2 As shown, Figure 2 This is a flowchart of a message storage method provided in an embodiment of this application.

[0028] Step 1011: The department sends a message entry application.

[0029] Step 1012: Whether the message approval is approved.

[0030] In some embodiments, the electronic device may approve the message based on the pre-approved message approval rules, or the approver may approve the message through the electronic device.

[0031] In some embodiments, if the message is approved, step 203 is executed; if the message is not approved, step 201 is executed.

[0032] Step 1013: Add the message to the message database.

[0033] In some embodiments, after a message is entered into a message database, the electronic device can also classify the messages according to department type or message type and display the classified messages.

[0034] In some embodiments, electronic devices can also transmit messages such as policy documents, work reports, notices, and announcements.

[0035] In some embodiments, the electronic device can also establish a message reporting standard system, which includes: message reporting data standards, message reporting business standards, and message entry process specifications. For example, the message reporting data standards mainly specify standardized reporting requirements such as message classification, core keywords, content, timeliness, reporting department, and related link descriptions; the message reporting process specifications are used to review the submitted message content, and only messages that meet the requirements can be valid.

[0036] In some embodiments, electronic devices can train models using messages, extract keywords from messages to construct a knowledge graph of messages, and provide support and convenience for message retrieval.

[0037] In some embodiments, electronic devices can perform text recognition on messages in a message database, summarize message content items based on the messages, and further process the messages using natural language processing technology to obtain message summaries that can be directly understood and used by users. This allows users to quickly understand the key information conveyed in the messages and comprehend the core data related to various business areas.

[0038] Step 102: Obtain the target user's historical behavior data in processing the message.

[0039] In some embodiments, electronic devices collect, organize, and analyze departmental meeting minutes, notices, and announcements corresponding to the target user, based on the target user's job function, in order to obtain the target user's behavioral data.

[0040] In some embodiments, the behavioral data includes at least one of the following: the target user's historical comments on the message, the target user's historical messages viewed, the target user's historical message keywords retrieved, the target user's historical click data, and the target user's historical activity trajectory.

[0041] Step 103: Determine the user needs of the target user based on the behavioral data.

[0042] In some embodiments, determining the user needs of the target user based on the behavioral data includes: determining the user needs corresponding to the behavioral data through a pre-trained intent recognition model; Before determining the user demand corresponding to the behavioral data through a pre-trained intent recognition model, the method further includes: The behavioral data is input into the intent recognition model to obtain the first output result; Add noise to the first output result to obtain the second output result; Based on the first output result, the second output result, and the preset output result, the cross-entropy value and the relative entropy value corresponding to the intent recognition model are determined; the cross-entropy value is used to determine the difference between the first output result and the preset output result, and the relative entropy value is used to determine the difference between the second output result and the preset output result; The intent recognition model is trained based on the cross-entropy value and the relative entropy value until the cross-entropy value is less than a second preset threshold and / or the relative entropy value is less than a third preset threshold, thereby obtaining the pre-trained intent recognition model.

[0043] Understandably, by determining the cross-entropy and relative entropy values ​​corresponding to the intent recognition model through the first output result, the second output result, and the preset output result, a more accurate intent recognition model can be trained, which is convenient for determining user needs through the intent recognition model in the future.

[0044] The following is a concrete example to illustrate this. An intent recognition model can consist of multiple Bidirectional Encoder Representations from Transformers (BERT) models, trained using large-scale unlabeled corpora, and outputting text semantics containing rich semantic information after training. Training the model using large-scale unlabeled corpora refers to the process of training a machine learning model (such as the BERT model) using massive amounts of unannotated text data.

[0045] In the embodiments of this application, the electronic device can set different BERT models according to different user types. For example, the electronic device can set a BERT model for users in different departments, users at different ends of an industry, and users in different fields. After setting the BERT model on the electronic device, behavioral data can be input into each BERT model to obtain the first output result corresponding to each BERT model.

[0046] To improve the model's robustness against interference and noise, electronic devices can add noise to the first output of each BERT model and use adversarial training to train the model, thereby enhancing its noise resistance. Optionally, electronic devices can add noise to the output layer of each BERT model; for example, they can add Gaussian noise to the first output of each BERT model to obtain the second output of each BERT model.

[0047] The electronic device then determines the cross-entropy and relative entropy values ​​of the entire intent recognition model based on the first output results, second output results, and preset output results of each BERT model. Based on these cross-entropy and relative entropy values, the model loss of the intent recognition model composed of multiple BERT models is calculated, thereby training the intent recognition model. Specifically, the electronic device can calculate the cross-entropy and relative entropy values ​​of the first output results, second output results, and preset output results through field matching. For example, the electronic device can match the fields of the first output results with the fields of the preset output results to obtain the cross-entropy value.

[0048] The preset output result represents the expected result under ideal conditions after inputting behavioral data into the intent recognition model. The cross-entropy value represents the difference between the first output result and the preset output result when the intent recognition model is undisturbed. A smaller cross-entropy value indicates higher accuracy of the output result when undisturbed, indicating that the intent recognition model has completed pre-training. Conversely, a larger cross-entropy value indicates lower accuracy of the output result when undisturbed, indicating that the intent recognition model has not completed pre-training and is currently in an abnormal state. The relative entropy value represents the difference between the second output result and the preset output result when the intent recognition model is disturbed. A smaller relative entropy value indicates higher accuracy of the output result when disturbed, indicating that the intent recognition model has completed pre-training. Conversely, a larger relative entropy value indicates lower accuracy of the output result when disturbed, indicating that the intent recognition model has not completed pre-training and is currently in an abnormal state.

[0049] Understandably, electronic devices can train an intent recognition model based on cross-entropy and relative entropy values ​​until both values ​​meet preset conditions, thus obtaining a pre-trained intent recognition model. For example, the intent recognition module completes training when the cross-entropy value is less than a second preset threshold and / or the relative entropy value is less than a third preset threshold.

[0050] Step 104: Based on the user requirements, generate personalized message information corresponding to the target user.

[0051] In some embodiments, the electronic device can generate personalized message information corresponding to the user's needs based on user requirements and messages in a message database. The personalized message information includes at least one of message annotations and message summaries, wherein the message annotations are generated based on the message summaries.

[0052] In other embodiments, when the personalized message information is an annotation, the electronic device can also generate comparative interpretations and suggestive annotations based on behavioral data, such as the user's historical annotation information, so that the user can quickly approve the message.

[0053] Understandably, in this embodiment, the electronic device helps users process messages efficiently and reduces the effort and time spent reading messages by generating message annotations and / or message summaries.

[0054] In some embodiments, the process by which an electronic device generates personalized message information may be: Generate at least one candidate personalized message based on the user's requirements; The at least one candidate personalized message is sorted based on the word-limited reward SBWR algorithm, and the candidate personalized message located at the preset position is taken as the personalized message corresponding to the target user.

[0055] For example, electronic devices can use word embedding and encoder-decoder structure transducer neural frameworks to achieve personalized message information output for messages.

[0056] Optionally, during the message decoding process, the electronic device can generate at least one candidate personalized message based on user needs, allowing the device to select the personalized message corresponding to the target user. The electronic device then sorts the at least one candidate personalized message using a word-limited reward (SBWR) algorithm, selecting the candidate personalized message at a preset position as the personalized message corresponding to the target user. For example, the preset position could be the first position in the sorting; however, this embodiment does not impose any restrictions on the preset position. It is understood that through this method, the electronic device can generate the most suitable personalized message for the user, helping them process messages efficiently and reducing the effort and time spent reading messages.

[0057] Electronic devices can employ different strategies simultaneously to expand the search space for candidate personalized messages. For example, they can use a best-first search algorithm to broaden the search space and generate more diverse candidate personalized messages based on behavioral data corresponding to the target user. This provides the target user with personalized messages of different styles or emphases. For instance, electronic devices can generate multiple candidate personalized messages based on behavioral data such as the target user's industry, region, level of government department, administrative agency, functional department of enterprise / institution, and user attributes.

[0058] In some embodiments, the electronic device may further perform step 101, i.e., after the messages are stored in the message database, use natural language processing technology to extract summaries from each message stored in the database to obtain initial summaries corresponding to each message. The electronic device then uses natural language generation technology to further generate corresponding personalized message information based on user needs from the initial summaries corresponding to each message.

[0059] In some embodiments, electronic devices can filter and generate personalized information summaries corresponding to target users from the initial summaries corresponding to each message through keyword matching, similarity calculation, and other methods.

[0060] In some embodiments, when the user is a first-time user or the user has no historical message processing behavior data, the electronic device can directly generate message summaries or message annotations for the user. However, the message summaries or message annotations are not personalized based on the user's behavior data, but are initial, general message summaries or message annotations obtained by summarizing and refining the various messages stored in the database.

[0061] Through the embodiments described in steps 101-104 above, the electronic device can acquire historical message processing behavior data of the target user, determine the target user's needs based on the behavior data, and then generate personalized message information corresponding to the target user based on the user's needs. Personalized message information helps users quickly extract the core points of a message, obtain important information from the message, and ensure that users can efficiently grasp and use the core information in the message, greatly improving the efficiency and quality of information absorption. This allows users to process messages efficiently, reducing the effort and time spent reading messages.

[0062] In some embodiments, when the target message includes numerical data corresponding to time data, the numerical data is predicted to determine the numerical data of the target time, and the numerical data of the target time is displayed, wherein the target message is the message corresponding to the personalized message information.

[0063] The following is a specific embodiment for illustration. Time data can refer to time-related data, such as 2025, August, the first quarter, the first cycle, etc. The numerical data corresponding to the time data can also refer to time-related numerical data, such as an annual GDP growth rate of 25% in 2025. In this case, 2025 is the time data, and 25% is the corresponding numerical data. The electronic device can predict the numerical data corresponding to the time data to determine the numerical data for the target time. The target time can be the previous time or the next time in the target message, and is not limited in this embodiment. For example, if the electronic device predicts an annual GDP growth rate of 25% for 2025 and obtains a predicted annual GDP growth rate of 30% for 2026, then 2026 is the target time, and 30% is the numerical data for the target time.

[0064] Understandably, in this embodiment, the electronic device can obtain the predicted numerical data of the target time by predicting the numerical data corresponding to the time data, so that the user does not need to determine the numerical data of the target time through calculation or prediction, saving the user's time and improving the user's efficiency.

[0065] In some embodiments, the numerical data for displaying the target time includes: The personalized message information displays numerical data of the target time.

[0066] For example, when the personalized message information is a message summary, the electronic device can highlight the numerical data predicting the target time in the message summary. Or, when the personalized message information is a message annotation, the electronic device can display the numerical data predicting the target time in the message annotation.

[0067] Understandably, in this embodiment, the electronic device displays numerical data of the target time in the personalized message information, which can help users quickly find the numerical data of the predicted target time and improve the user experience.

[0068] In mobile work scenarios, users may need to work in different environments and conditions, such as during commutes, during meeting breaks, or when multitasking. Traditional methods of reading reports may be limited by time and space. In some embodiments, electronic devices can convert personalized message information into audio files and play the audio files.

[0069] Optionally, electronic devices can integrate Conversational Text-to-Speech (ChatTTS) technology to convert personalized messages into natural and fluent speech files. ChatTTS is an open-source speech generation model specifically designed for conversational scenarios, particularly suitable for dialogue tasks in large language model assistants, as well as applications such as conversational audio and video introductions. Supporting both Chinese and English, and trained using approximately 100,000 hours of Chinese and English data, ChatTTS demonstrates high quality and naturalness in speech synthesis.

[0070] Optionally, electronic devices can use ChatTTS technology to mimic real human voices, and can even adjust the playback speed, tone, and intonation according to specific scenarios or preferences. The voice files generated by the electronic devices can also be saved, and high-quality output audio is ensured through integrated noise reduction, equalization, and compression algorithms. Electronic devices can also utilize cloud infrastructure to store and transmit audio files, ensuring low latency and high availability. Understandably, by converting personalized message information into voice files and playing them, electronic devices allow users to listen to personalized messages anytime, anywhere, without the need for manual reading.

[0071] In some embodiments, the electronic device can adjust the speech rate and / or tone of the voice file corresponding to the personalized message information based on the type of the target message. Optionally, ChatTTS technology can not only convert traditional written messages into speech, but also make the broadcast content more vivid and closer to actual work scenarios by simulating specific situations and tones. This broadcasting method uses emotional speech synthesis, and the electronic device can adjust the speech rate, tone, and emotional color according to the type of message (such as formal meeting minutes, emergency notices, daily work reports, etc.). It can also be manually adjusted and adapted according to actual usage, thereby creating a sense of context as if listening to a report on-site. For example, when broadcasting urgent matters, the voice played by the electronic device will be more urgent and serious; while when conveying daily progress, the voice played by the electronic device will adopt a relaxed and calm tone. This voice broadcasting method achieves immersion and participation in information delivery, allowing the audience to be psychologically closer to the actual application scenario of the message, thereby improving the absorption rate and depth of understanding of the information.

[0072] When users spend long periods immersed in large amounts of text reports, it often puts pressure on their eyesight, leading to visual fatigue. ChatTTS technology's broadcast function, by converting text information into audio, allows users to receive information without looking directly at the screen, significantly reducing eye strain. Simultaneously, audio broadcasting helps users maintain focus and reduces distractions caused by visual fatigue. Especially after prolonged visual work, using hearing to acquire information allows the eyes to rest without interfering with the work process.

[0073] Secondly, through voice broadcasting, users can listen to work reports while engaging in daily activities such as commuting, exercising, or doing housework, effectively saving reading time and improving work efficiency. ChatTTS technology combines natural language processing and deep learning algorithms to accurately capture key points in the text and present them in a clear and fluent audio format, ensuring that users can easily understand the content.

[0074] The above embodiments provide users with convenient information access channels, comprehensively assisting their operations, enabling information summarization and consolidation, strengthening guidance and command of assigned tasks, saving labor costs, and improving work efficiency and quality. Furthermore, the broadcast of messages allows users to understand key work points anytime, anywhere, achieving efficient mastery of critical information. In addition, these embodiments can also positively impact business concepts and workflows, greatly enhancing enterprises' informatization level, management level, and comprehensive service capabilities within the industry, effectively improving supply chain support and increasing domestic production levels.

[0075] If the implementation process of the above embodiments of this application is illustrated in publicly available materials such as product white papers, official websites, product manuals, product posters, user manuals, product UI, physical products, or photographs, then it proves that the solution of this application has been used. Alternatively, it can be determined whether the software source code has the same method, parameters, or architectural style as this application. Alternatively, it can be determined whether the above embodiments of this application have been used through software operation results (including software interface, operation input parameters, prompt information, etc.). Alternatively, it can be determined whether the above embodiments of this application have been used through testing techniques such as constructing simulated scenarios, simulating user usage, and performing packet capture analysis on data input and output. Alternatively, it can be determined whether the above embodiments of this application have been used through reverse engineering methods such as disassembling hardware, cracking software, simulating software functions, and analyzing code.

[0076] Please refer to Figure 3 , Figure 3 This is a schematic diagram of an information generation device 300 according to an embodiment of this application. The information generation device 300 includes: Construction module 301 is used to construct a message database, which is used to store messages; The acquisition module 302 is used to acquire historical behavior data of the target user in processing the message; The determining module 303 is used to determine the user needs of the target user based on the behavioral data; The generation module 304 is used to generate personalized message information corresponding to the target user based on the user requirements.

[0077] Optionally, the personalized message information includes at least one of message annotation and message summary, wherein the message annotation is generated based on the message summary.

[0078] Optionally, the generation module 304 can also be used for: Generate at least one candidate personalized message based on the user's requirements; The at least one candidate personalized message is sorted based on the word-limited reward SBWR algorithm, and the candidate personalized message located at the preset position is taken as the personalized message corresponding to the target user.

[0079] Optionally, the behavioral data includes at least one of the following: the target user's historical comments on the message, the target user's historical messages viewed, the target user's historical message keywords retrieved, the target user's historical click data, and the target user's historical activity trajectory; The determining module 303 can also be used for: The user's needs corresponding to the behavioral data are determined by a pre-trained intent recognition model; Before determining the user demand corresponding to the behavioral data through a pre-trained intent recognition model, the method further includes: The behavioral data is input into the intent recognition model to obtain the first output result; Add noise to the first output result to obtain the second output result; Based on the first output result, the second output result, and the preset output result, the cross-entropy value and the relative entropy value corresponding to the intent recognition model are determined; the cross-entropy value is used to determine the difference between the first output result and the preset output result, and the relative entropy value is used to determine the difference between the second output result and the preset output result; The intent recognition model is trained based on the cross-entropy value and the relative entropy value until the cross-entropy value is less than a second preset threshold and / or the relative entropy value is less than a third preset threshold, thereby obtaining the pre-trained intent recognition model.

[0080] Optionally, the information generation device 300 may further include: The prediction module is used to predict the numerical data when the target message includes numerical data corresponding to time data, so as to determine the numerical data of the target time and display the numerical data of the target time, wherein the target message is the message corresponding to the personalized message information.

[0081] Optionally, the prediction module may also include: Display submodule: Displays the numerical data of the target time in the personalized message information.

[0082] Optionally, the information generation device 300 may further include: The playback module is used to convert the personalized message information into an audio file and play the audio file.

[0083] The information generation apparatus 300 provided in this application embodiment can perform the above-described... Figure 1 The method embodiments shown are similar in principle and technical effect, and will not be described again here.

[0084] This application also provides an electronic device. Since the principle by which this electronic device solves the problem is similar to the information generation method in the embodiments of this application, the implementation of this electronic device can be found elsewhere. Figure 1 The implementation of the method shown will not be repeated here. Figure 4 As shown, the electronic device according to an embodiment of this application includes: a processor 410, configured to read a program from a memory 420 and execute the following processes: Build a message database, which is used to store messages; Obtain behavioral data on the target user's historical message processing. Determine the user needs of the target users based on behavioral data; Based on user needs, generate personalized message information corresponding to the target user.

[0085] Optionally, the processor 410 is also used to read the program from the memory 420 and perform the following steps: Generate at least one candidate personalized message based on user needs; Based on the limited word reward SBWR algorithm, at least one candidate personalized message is sorted, and the candidate personalized message located at the preset position is taken as the personalized message corresponding to the target user.

[0086] Optionally, the processor 410 is also used to read the program from the memory 420 and perform the following steps: Determining target user needs based on behavioral data includes: using a pre-trained intent recognition model to determine the user needs corresponding to the behavioral data; Before determining the user needs corresponding to the behavioral data through a pre-trained intent recognition model, the method also includes: Input the behavioral data into the intent recognition model to obtain the first output result; Add noise to the first output to obtain the second output; Based on the first output result, the second output result, and the preset output result, the cross-entropy value and the relative entropy value corresponding to the intent recognition model are determined; the cross-entropy value is used to determine the difference between the first output result and the preset output result, and the relative entropy value is used to determine the difference between the second output result and the preset output result. The intent recognition model is trained based on the cross-entropy value and the relative entropy value until the cross-entropy value is less than the second preset threshold and / or the relative entropy value is less than the third preset threshold, thus obtaining the pre-trained intent recognition model.

[0087] Optionally, the processor 410 is also used to read the program from the memory 420 and perform the following steps: If the target message includes numerical data corresponding to time data, the numerical data is predicted to determine the numerical data for the target time, and the numerical data for the target time is displayed. Here, the target message is the message corresponding to the personalized message information.

[0088] Optionally, the processor 410 is also used to read the program from the memory 420 and perform the following steps: Numerical data displaying the target time includes: Display numerical data of the target time in the personalized message information.

[0089] Optionally, the processor 410 is also used to read the program from the memory 420 and perform the following steps: Convert personalized message information into an audio file and play the audio file.

[0090] Among them, Figure 4 In this context, the bus architecture can include any number of interconnected buses and bridges, specifically linking various circuits of one or more processors represented by processor 410 and memory represented by memory 420 together. The bus architecture can also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. The bus interface provides the interface.

[0091] The electronic device provided in this application embodiment can perform the above-described functions. Figure 1 The method embodiments shown are similar in principle and technical effect, and will not be described again here.

[0092] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the above-described... Figure 1 The various processes in the embodiments of the information generation method can achieve the same technical effect, and will not be described again here to avoid repetition. The computer-readable storage medium used is such as read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0093] This application also provides a computer program / program product, which is stored in a storage medium and executed by at least one processor to implement the above. Figure 1 The various processes in the embodiments of the information generation method can achieve the same technical effect, and will not be described again here to avoid repetition.

[0094] In the several embodiments provided in this application, it should be understood that the disclosed methods and apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0095] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can be physically included separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0096] The integrated units implemented as software functional units described above can be stored in a computer-readable storage medium. These software functional units, stored in a storage medium, include several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute some steps of the transmission and reception methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] The above are preferred embodiments of this application. It should be noted that, for those skilled in the art, several improvements and modifications can be made without departing from the principles of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. An information generation method characterized by comprising: The method comprises the following steps: constructing a message database for storing messages; obtaining behavior data of a target user in processing the messages; determining user demand of the target user according to the behavior data; generating personalized message information corresponding to the target user based on the user demand.

2. The method of claim 1, wherein: the personalized message information comprises at least one of a message annotation and a message abstract, and the message annotation is generated based on the message abstract.

3. The method of claim 1, wherein, The method further comprises the following steps: generating at least one candidate personalized message information based on the user demand; sorting the at least one candidate personalized message information based on a limited word reward SBWR algorithm, and taking the candidate personalized message information located at a preset position as the personalized message information corresponding to the target user.

4. The method of claim 1, wherein, The behavior data comprises at least one of historical annotations of the target user on the messages, historical messages browsed by the target user, historical keywords of the messages searched by the target user, historical click data of the target user, and historical activity tracks of the target user. The method further comprises the following steps: inputting the behavior data into the intent recognition model to obtain a first output result; adding noise to the first output result to obtain a second output result; determining a cross-entropy value and a relative entropy value corresponding to the intent recognition model based on the first output result, the second output result, and a preset output result; the cross-entropy value is used to determine a difference value between the first output result and the preset output result, and the relative entropy value is used to determine a difference value between the second output result and the preset output result; training the intent recognition model based on the cross-entropy value and the relative entropy value until the cross-entropy value is less than a second preset threshold value and / or the relative entropy value is less than a third preset threshold value, to obtain the pre-trained intent recognition model. The method further comprises the following steps:

5. The method according to any one of claims 1-4, characterized in that, in a case where a target message comprises numerical data corresponding to time data, predicting the numerical data to determine numerical data of a target time, and displaying the numerical data of the target time, wherein the target message is a message corresponding to the personalized message information. The method further comprises the following steps:

6. The method of claim 5, wherein, displaying the numerical data of the target time in the personalized message information. The method further comprises the following steps:

7. The method of claim 1, wherein, converting the personalized message information into a voice file, and playing the voice file. The method comprises the following steps:

8. An information generation apparatus characterized by comprising: constructing a message database for storing messages; obtaining behavior data of a target user in processing the messages; determining user demand of the target user according to the behavior data; ​ The generating module is configured to generate the personalized message information corresponding to the target user based on the user demand.

9. An electronic device, comprising: The information generation method comprises the following steps: The processor, the memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps in the information generation method according to any one of claims 1 to 7.

10. A computer readable storage medium for storing a computer program, characterized in that, The computer program, when executed by the processor, implements the steps in the information generation method according to any one of claims 1 to 7.

11. A computer program product, characterised in that, The computer program, when executed by the processor, implements the steps in the information generation method according to any one of claims 1 to 7.

Images