Knowledge processing method and system, storage medium and program product

Abstract

The invention provides a knowledge processing method and system, a storage medium and a program product, and relates to the field of artificial intelligence. The method comprises the following steps: constructing a user knowledge graph according to historical operation of each user on knowledge in a public knowledge base; processing the user knowledge graph to obtain a user feature vector and a knowledge feature vector; based on the user feature vector and the knowledge feature vector, obtaining an embedded vector matrix of multiple pieces of candidate knowledge; and based on the historical operation of the target user on the knowledge and the embedding vector matrix of the plurality of candidate knowledge, performing knowledge recommendation on the target user.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence, and in particular to a knowledge processing method, system, storage medium and program product. Background Art

[0002] With the rapid advancement of artificial intelligence technology, large language models have become the key technical cornerstone in many fields, especially in intelligent question-answering systems, which are changing the way information is acquired and processed with unprecedented capabilities. In particular, combined with advanced retrieval enhancement generation frameworks, government affairs intelligent question-answering is particularly prominent among many application scenarios, and how to accurately recommend government affairs knowledge has become particularly important. Summary of the invention

[0003] A technical problem to be solved by the present disclosure is to provide a knowledge processing method, system, storage medium and program product, which can improve the accuracy of knowledge recommendation.

[0004] According to one aspect of the present disclosure, a knowledge processing method is proposed, including: constructing a user knowledge graph based on each user's historical operations on knowledge in a public knowledge base; processing the user knowledge graph to obtain a user feature vector and a knowledge feature vector; obtaining an embedding vector matrix of multiple candidate knowledge based on the user feature vector and the knowledge feature vector; and recommending knowledge to a target user based on the target user's historical operations on knowledge and the embedding vector matrix of multiple candidate knowledge.

[0005] In some embodiments, in response to the target user performing at least one of an editing operation and an updating operation on the recommended knowledge, the public knowledge base is updated.

[0006] In some embodiments, based on the target user's historical operations on knowledge and the embedding vector matrices of multiple candidate knowledge, knowledge recommendation to the target user includes: performing information enhancement on the target user's historical operations on knowledge to obtain a first embedding vector matrix corresponding to the target user; performing self-attention processing on the first embedding vector matrix to obtain a second embedding vector matrix; using the second embedding vector matrix and the embedding vector matrices of multiple candidate knowledge to score each of the multiple candidate knowledge; and based on the score of each candidate knowledge, recommending knowledge to the target user.

[0007] In some embodiments, information enhancement is performed on the target user's historical operations on knowledge to obtain a first embedding vector matrix corresponding to the target user, including: averaging multiple contexts of each session in the target user's historical operations on knowledge to obtain a first fused embedding vector representation; performing time-varying information fusion on each session in the target user's historical operations on knowledge to obtain a second fused embedding vector representation; performing weighted calculation on the first fused embedding vector representation and the second fused embedding vector representation to obtain a third fused embedding vector representation corresponding to each session; concatenating the third fused embedding vector representation corresponding to each session and the time of each session, and processing them using a feedforward neural network to obtain an embedding vector matrix corresponding to each session; and obtaining a first embedding vector matrix corresponding to the target user based on the embedding vector matrix corresponding to each session.

[0008] In some embodiments, performing self-attention processing on the first embedding vector matrix to obtain a second embedding vector matrix includes: using the query parameter matrix, key parameter matrix and value parameter matrix of the self-attention algorithm to perform normalized exponential processing on the first embedding vector matrix to obtain a second embedding vector matrix.

[0009] In some embodiments, processing the user knowledge graph to obtain the user feature vector and the knowledge feature vector includes: using a bipartite graph to process the user knowledge graph to obtain the user feature vector and the knowledge feature vector.

[0010] In some embodiments, updating the public knowledge base further includes: in response to the recommended knowledge being reviewed and approved, forming public knowledge; and updating the public knowledge base using the public knowledge.

[0011] In some embodiments, in response to the recommended knowledge being reviewed and approved, forming public knowledge includes: classifying the recommended knowledge; and forming public knowledge after the recommended knowledge is reviewed and approved according to its category.

[0012] In some embodiments, the knowledge processing method also includes: distributed collection of original knowledge; data conversion and preprocessing of the original knowledge; knowledge extraction of the original knowledge that has undergone data conversion and preprocessing around the business domain to form a structured knowledge representation; and organizing the structured knowledge representation to obtain a public knowledge base.

[0013] According to another aspect of the present disclosure, a knowledge processing system is also proposed, including a knowledge recommendation module, which includes: a construction unit, configured to construct a user knowledge graph according to each user's historical operations on knowledge in a public knowledge base; a processing unit, configured to process the user knowledge graph to obtain a user feature vector and a knowledge feature vector; a candidate unit, configured to obtain an embedding vector matrix of multiple candidate knowledge based on the user feature vector and the knowledge feature vector; and a recommendation unit, configured to recommend knowledge to a target user based on the target user's historical operations on knowledge using the user feature vector and the knowledge feature vector.

[0014] In some embodiments, the knowledge processing system further includes: a sharing module configured to update the public knowledge base in response to at least one of an editing operation and an updating operation performed by the target user on the recommended knowledge.

[0015] In some embodiments, the knowledge processing system also includes: a collection module, which is configured to collect original knowledge in a distributed manner, perform data conversion and preprocessing on the original knowledge, extract knowledge from the original knowledge that has undergone data conversion and preprocessing around the business domain to form a structured knowledge representation, and organize the structured knowledge representation to obtain a public knowledge base.

[0016] According to another aspect of the present disclosure, a knowledge processing system is also proposed, including: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the above-mentioned knowledge processing method based on instructions stored in the memory.

[0017] According to another aspect of the present disclosure, a computer-readable storage medium is also provided, on which computer program instructions are stored, and when the instructions are executed by a processor, the knowledge processing method as described above is implemented.

[0018] According to another aspect of the present disclosure, a computer program product is also proposed, including a computer program or instructions, and the computer program or instructions implement the above-mentioned knowledge processing method when executed by a processor.

[0019] Other features and advantages of the present disclosure will become apparent from the following detailed description of exemplary embodiments of the present disclosure with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings, which constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0021] The present disclosure may be more clearly understood from the following detailed description with reference to the accompanying drawings, in which:

[0022] Figure 1 Schematic diagram of some embodiments of the knowledge processing method disclosed in the present invention;

[0023] Figure 2 Schematic diagrams of some embodiments of the knowledge processing flow disclosed herein;

[0024] Figure 3 Schematic diagram of the flow chart of other embodiments of the knowledge processing method disclosed in the present invention;

[0025] Figure 4 Schematic diagram of the flow chart of other embodiments of the knowledge processing method disclosed in the present invention;

[0026] Figure 5 It is a schematic diagram of the structure of some embodiments of the knowledge recommendation module of the knowledge processing system disclosed in the present invention;

[0027] Figure 6 It is a schematic diagram of the structure of some embodiments of the knowledge processing system disclosed in the present invention;

[0028] Figure 7 It is a schematic diagram of the structure of some other embodiments of the knowledge processing system disclosed in the present invention;

[0029] Figure 8 It is a schematic diagram of the structure of some embodiments of the government knowledge governance system disclosed in the present invention. DETAILED DESCRIPTION

[0030] Various exemplary embodiments of the present disclosure will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present disclosure unless otherwise specifically stated.

[0031] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0032] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.

[0033] Technologies, methods, and equipment known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, methods, and equipment should be considered as part of the specification.

[0034] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0035] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0036] In order to make the objectives, technical solutions and advantages of the present disclosure more clearly understood, the present disclosure is further described in detail below in combination with specific embodiments and with reference to the accompanying drawings.

[0037] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, storage, and sharing of user personal information involved in the technical solution of this disclosure are in compliance with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken for user personal information to prevent illegal access to user personal information data and maintain the security of user personal information, network security, and national security.

[0038] Figure 1 1 is a flowchart of some embodiments of the knowledge processing method disclosed in the present invention, and the embodiments include steps S11-S14.

[0039] In step S11, a user knowledge graph is constructed based on each user's historical operations on the knowledge in the public knowledge base.

[0040] This knowledge is, for example, public knowledge such as government affairs knowledge. Government affairs knowledge mainly refers to question-answer pairs transformed from business personnel, policy documents, and structured information of matters, for example, including questions and answers.

[0041] In some embodiments, a user-knowledge graph is constructed based on the salesperson's historical operations on the question-answer pair. Figure 2 As shown, Figure 2 The schematic diagram of some embodiments of the knowledge processing flow of the present disclosure is as follows: i Indicates that the question and answer are k n Indicates that the weight k i,n It is 0 or 1, 1 means the user has clicked on the question and answer pair, 0 means the user has not clicked on the question and answer pair, and the constructed user-knowledge graph is represented by G.

[0042] In step S12, the user knowledge graph is processed to obtain a user feature vector and a knowledge feature vector.

[0043] In some embodiments, the user knowledge graph is processed using a bipartite graph to obtain a user feature vector and a knowledge feature vector.

[0044] Bipartite graphs can fully model the user-knowledge interaction. Using bipartite graphs for learning preserves the second-order proximity, making its conditional probability close to its empirical distribution, and obtaining invisible user features and knowledge features.

[0045] User characteristics include, for example, the user's position, business attributes, interaction behavior, etc.

[0046] In step S13, based on the user feature vector and the knowledge feature vector, an embedding vector matrix of multiple candidate knowledge is obtained.

[0047] In some embodiments, user feature vectors and knowledge feature vectors are recalled according to a predetermined number to obtain an embedding vector matrix of multiple candidate knowledge.

[0048] For example, recall is performed based on the learned user feature vector and knowledge feature vector, and for example, the recall pool is initially set to 50. Those skilled in the art should understand that the setting of the recall pool here is only for example and can be set according to actual conditions.

[0049] In step S14, knowledge is recommended to the target user based on the target user's historical operations on knowledge and the embedding vector matrices of multiple candidate knowledge.

[0050] For example, the softmax algorithm is used to score candidate knowledge, and then knowledge is recommended to target users according to the score.

[0051] In the above embodiment, by constructing a user knowledge graph, a high-quality embedding vector matrix of multiple candidate knowledge is obtained, and then based on the target user's historical operations on knowledge and the embedding vector matrix of multiple candidate knowledge, personalized knowledge recommendations are provided to the target user, thereby achieving accurate knowledge recommendations and improving service quality and efficiency.

[0052] In some embodiments of the present disclosure, step S14, based on the historical operations of the target user on the knowledge and the embedding vector matrix of multiple candidate knowledge, the implementation of knowledge recommendation for the target user is as follows: Figure 3 As shown, steps S141-S144 are included.

[0053] In step S141, information enhancement is performed on the target user's historical operations on knowledge to obtain a first embedding vector matrix corresponding to the target user.

[0054] like Figure 2 As shown, the target user u i The historical interaction of knowledge pairs is represented by a vector: Then, through the context enhancement layer, the first embedding vector matrix is ​​obtained

[0055] This step fuses the periodicity and contextual fine-grained information of each session in historical operations, which can reduce the impact of time fluctuations and timestamp divisions on session embedding representation.

[0056] In some embodiments, multiple contexts of each session in the target user's historical operations on knowledge are averaged to obtain a first fused embedding vector representation; time-varying information is fused for each session in the target user's historical operations on knowledge to obtain a second fused embedding vector representation; weighted calculation is performed on the first fused embedding vector representation and the second fused embedding vector representation to obtain a third fused embedding vector representation corresponding to each session; the third fused embedding vector representation corresponding to each session and the time of each session are concatenated and processed using a feedforward neural network to obtain an embedding vector matrix corresponding to each session; based on the embedding vector matrix corresponding to each session, the first embedding vector matrix corresponding to the target user is obtained.

[0057] like Figure 2 As shown, according to Get multiple operation records of the target user (u i , c j , k n , t j ), (u i , c j , k n , t j ) represents user u i At time t j For category c j Knowledge n For each session The embedding vector representation is obtained by averaging multiple contexts The time-varying information of each session is fused to obtain the embedded vector representation ts is a timestamp. The time-varying properties here include the basic characteristics of time, including year, month, day, hour, minute, and second, and the time period characteristics, including hours in a day, days of the week in a month, and months in a year.

[0058] Through a balanced context- and time-aware information augmentation paradigm, each session The expression α is used to balance the weight of contextual information and time-varying information. Applying the above information enhancement strategy, the third fusion embedding vector representation can be obtained: Finally, according to ts aggregation, we get Among them, FFN is a feed-forward neural network, and Concat represents the concatenation operation of the embedded vector.

[0059] For target user u i A set of historical operations The embedding vector matrix E( hist ) = {a1, a2, ..., a n} to indicate.

[0060] In step S142, self-attention processing is performed on the first embedding vector matrix to obtain a second embedding vector matrix.

[0061] In this step, the self-attention mechanism can effectively capture the sequential dependencies between operation records in user interactions. For example, the embedding after the self-attention layer is expressed as E′ (hist) ={a′1,a′2,...,a′ n}.

[0062] In some embodiments, the first embedding vector matrix is ​​normalized and indexed using the query parameter matrix, key parameter matrix, and value parameter matrix of the self-attention algorithm to obtain a second embedding vector matrix.

[0063] like Figure 2 As shown in the figure, history represents the historical operations of the target user, which can be used to construct the query parameter matrix, key parameter matrix and value parameter matrix of the self-attention algorithm, and then the query parameter matrix, key parameter matrix and value parameter matrix are used to perform normalized index processing on the first embedding vector matrix to obtain the second embedding vector matrix E′ (hist) ={a′1,a′2,...,a′ n}.

[0064] In step S143, each of the plurality of candidate knowledge is scored using the second embedding vector matrix and the embedding vector matrices of the plurality of candidate knowledge.

[0065] In some embodiments, the embedding vector matrix of multiple candidate knowledge can be expressed as Use the softmax algorithm to score the candidate knowledge Among them, each It indicates the score that the model gives to the knowledge pair ki. The higher the score, the more likely the model thinks that the user is likely to operate on this knowledge pair next.

[0066] For example, using the formula The score of the candidate knowledge is obtained, where Sum(·) represents the weighted sum of the last dimension, and d is determined by parameter experiments. Those skilled in the art should understand that the scoring of subsequent knowledge can be implemented using a variety of algorithms, and the above formula is only used as an example.

[0067] In step S144, knowledge is recommended to the target user based on the score of each candidate knowledge.

[0068] For example, the candidate knowledge with the highest scores is recommended to the target user. For another example, the top-K (the k with the highest scores) knowledge is recommended to the salesperson in order of scores.

[0069] In the above embodiment, context-enhanced conversation recommendation is adopted to improve the accuracy of knowledge recommendation.

[0070] In some embodiments of the present disclosure, in response to the target user performing at least one of an editing operation and an updating operation on the recommended knowledge, the public knowledge base is updated.

[0071] For example, with the development of business systems, the speed of information update is accelerated, and new knowledge is constantly emerging and accumulating. However, the traditional government knowledge governance mechanism is often unable to keep up with these changes in a timely manner, resulting in the inability of large-scale model-driven intelligent question-answering systems to provide the latest information. In this embodiment, users can edit or update the recommended knowledge, so that the knowledge in the public knowledge base can be the latest knowledge, and the knowledge in the public knowledge base can be shared.

[0072] In some embodiments, in response to the recommended knowledge being reviewed and approved, public knowledge is formed; and the public knowledge base is updated using the public knowledge.

[0073] The construction of a knowledge base usually relies on a large number of data sources, including office guides, tutorials, books, articles, etc. However, the quality of these data sources varies and may contain erroneous and biased information. Ensuring the reliability of these data sources is a huge challenge. In this embodiment, the recommended knowledge is reviewed to improve the quality of the knowledge, thereby improving the accuracy of the knowledge recommendation.

[0074] For example, for the knowledge collected and fed back, an effective collaboration mechanism between artificial intelligence and human experts has been established to ensure the quality of knowledge. For example, the accuracy of the knowledge automatically entered by the machine or the recommended knowledge needs to be reviewed and calibrated by experts, and government affairs knowledge is formed after the review.

[0075] In some embodiments, the recommended knowledge is classified; after the recommended knowledge is reviewed and approved according to its category, it forms public knowledge.

[0076] In this embodiment, different categories of knowledge are reviewed according to their categories, which can improve the quality of knowledge on the one hand and ensure the authority of knowledge on the other hand.

[0077] In some embodiments of the present disclosure, the collection of public knowledge is the basis for the subsequent recommendation process. Therefore, the collection of data from multiple sources is also very important. In the context of multi-source government data, it is necessary to solve the problems of real-time and accuracy of data collection. Figure 4 As shown, Figure 4 Schematic diagram of the flow chart of other embodiments of the knowledge processing method disclosed in the present invention, including steps S41-S44.

[0078] In step S41, original knowledge is collected in a distributed manner.

[0079] The original data may exist in multiple formats, including txt, word, excel, pdf, and json, covering six major types of data: policy documents, service guidelines, detailed operation instructions, training materials, question and answer records, and public letters.

[0080] In step S42, data conversion and preprocessing are performed on the original knowledge.

[0081] In some embodiments, raw data of various formats and structures are converted into a unified format and then pre-processed such as quality assessment and cleaning.

[0082] Specifically, the original data is processed by data conversion, data segmentation, stop word removal, feature construction, etc. Since the general segmentation is not very accurate for government affairs, the government affairs knowledge is divided based on the constructed government affairs dictionary.

[0083] In step S43, knowledge extraction is performed on the original knowledge that has undergone data conversion and preprocessing around the business domain to form a structured knowledge representation.

[0084] For example, around the government business domain, question-answer pairs are extracted for the matters to which they belong. The question-answer pair information includes the business domain, the question, and the corresponding answer. Through this process, valuable knowledge points can be extracted from the preprocessed data to form a structured knowledge representation.

[0085] In step S44, the structured knowledge representation is sorted to obtain a public knowledge base.

[0086] For example, on the basis of knowledge extraction, the core links of knowledge classification, labeling, storage, etc. are organized and summarized to form a knowledge system that is easy to retrieve and use.

[0087] In the process of using the knowledge base, knowledge from different sources is often scattered in different departments or business systems, which easily forms knowledge islands. These knowledge islands lack effective connection and integration, resulting in the system being unable to fully utilize these scattered knowledge resources. In this embodiment, through the distributed knowledge acquisition pipeline, automatic structured processing and segmentation of large-scale data, and knowledge extraction can ensure data consistency and redundancy in a distributed environment.

[0088] Figure 5 It is a schematic diagram of the structure of some embodiments of the knowledge recommendation module of the knowledge processing system of the present disclosure. The knowledge recommendation module 500 includes: a construction unit 510, a processing unit 520, a candidate unit 530 and a recommendation unit 540.

[0089] The construction unit 510 is configured to construct a user knowledge graph according to each user's historical operations on the knowledge in the public knowledge base.

[0090] The knowledge is, for example, public knowledge such as government affairs knowledge.

[0091] The processing unit 520 is configured to process the user knowledge graph to obtain a user feature vector and a knowledge feature vector.

[0092] In some embodiments, the user knowledge graph is processed using a bipartite graph to obtain a user feature vector and a knowledge feature vector.

[0093] The candidate unit 530 is configured to obtain an embedding vector matrix of multiple candidate knowledge based on the user feature vector and the knowledge feature vector.

[0094] In some embodiments, user feature vectors and knowledge feature vectors are recalled according to a predetermined number to obtain an embedding vector matrix of multiple candidate knowledge.

[0095] The recommendation unit 540 is configured to recommend knowledge to the target user based on the target user's historical operations on knowledge by using the user feature vector and the knowledge feature vector.

[0096] In some embodiments, the recommendation unit 540 is configured to perform information enhancement on the target user's historical operations on knowledge to obtain a first embedding vector matrix corresponding to the target user; perform self-attention processing on the first embedding vector matrix to obtain a second embedding vector matrix; use the second embedding vector matrix and the embedding vector matrices of multiple candidate knowledge to score each of the multiple candidate knowledge; and recommend knowledge to the target user based on the score of each candidate knowledge.

[0097] Information enhancement is performed on the target user's historical operations on knowledge to obtain a first embedding vector matrix corresponding to the target user, including: averaging multiple contexts of each session in the target user's historical operations on knowledge to obtain a first fused embedding vector representation; performing time-varying information fusion on each session in the target user's historical operations on knowledge to obtain a second fused embedding vector representation; weighted calculation is performed on the first fused embedding vector representation and the second fused embedding vector representation to obtain a third fused embedding vector representation corresponding to each session; the third fused embedding vector representation corresponding to each session and the time of each session are concatenated and processed using a feedforward neural network to obtain an embedding vector matrix corresponding to each session; and a first embedding vector matrix corresponding to the target user is obtained based on the embedding vector matrix corresponding to each session.

[0098] Performing self-attention processing on the first embedding vector matrix to obtain a second embedding vector matrix includes: using a query parameter matrix, a key parameter matrix, and a value parameter matrix of the self-attention algorithm to perform normalized exponential processing on the first embedding vector matrix to obtain a second embedding vector matrix.

[0099] In the above embodiment, by constructing a user knowledge graph, a high-quality embedding vector matrix of multiple candidate knowledge is obtained, and then based on the target user's historical operations on knowledge and the embedding vector matrix of multiple candidate knowledge, personalized knowledge recommendations are provided to the target user, thereby achieving accurate knowledge recommendations and improving service quality and efficiency.

[0100] In some embodiments of the present disclosure, Figure 6 As shown, Figure 6 This is a structural diagram of some embodiments of the knowledge processing system disclosed in the present invention. The knowledge processing system includes a knowledge recommendation module 500 and a sharing module 610, which is configured to update the public knowledge base in response to at least one of an editing operation and an updating operation performed by a target user on the recommended knowledge.

[0101] For example, the sharing module 610 is configured to form public knowledge in response to the recommended knowledge being reviewed and approved; and update the public knowledge base using the public knowledge.

[0102] In some embodiments, the recommended knowledge is classified; after the recommended knowledge is reviewed and approved according to its category, it forms public knowledge.

[0103] Through this sharing module, the knowledge in the public knowledge base can be kept updated and of higher quality, thereby improving the accuracy of knowledge recommendation.

[0104] In other embodiments, the knowledge processing system also includes a collection module 620, which is configured to collect original knowledge in a distributed manner, perform data conversion and preprocessing on the original knowledge, extract knowledge from the original knowledge that has undergone data conversion and preprocessing around the business domain to form a structured knowledge representation, organize the structured knowledge representation, and obtain a public knowledge base.

[0105] In this embodiment, the distributed knowledge acquisition pipeline is used to automatically structure and segment large-scale data, and extract knowledge, so as to ensure data consistency and redundancy in a distributed environment.

[0106] Figure 7Schematic diagram of the structure of some other embodiments of the knowledge processing system of the present disclosure, the knowledge processing system includes a memory 710 and a processor 720. Wherein: the memory 710 can be a disk, a flash memory or any other non-volatile storage medium. The memory is used to store the instructions in the above embodiments. The processor 720 is coupled to the memory 710 and can be implemented as one or more integrated circuits, such as a microprocessor or a microcontroller. The processor 720 is used to execute the instructions stored in the memory.

[0107] In some embodiments, the processor 720 is coupled to the memory 710 via a BUS 730. The knowledge processing system can also be connected to an external storage device 750 via a storage interface 740 to call external data, and can also be connected to a network or another computer system (not shown) via a network interface 760. No further detailed description will be given here.

[0108] The following will introduce the solution of this disclosure by taking the processing of government affairs knowledge as an example. The core challenge of the government affairs intelligent question-answering system is how to process and respond to policy interpretation and consulting services, which requires the system to have a high level of government affairs knowledge collection, update and sharing method capabilities.

[0109] Figure 8 It is a schematic diagram of the structure of some embodiments of the government knowledge governance system disclosed in the present invention.

[0110] The government knowledge governance system includes a distributed knowledge acquisition module 810, a feedback-based knowledge update and recommendation module 820, and a knowledge review and sharing module 830.

[0111] Distributed knowledge collection is the foundation of the government knowledge governance system. The distributed knowledge collection module 810 can perform data conversion, knowledge extraction and knowledge collection. For example, for different parts, corresponding government knowledge is collected separately. The distributed knowledge collection module 810 supports automatic structured processing and segmentation of large-scale data, knowledge extraction, and ensures data consistency and redundancy in a distributed environment.

[0112] The feedback-based knowledge update and recommendation module 820 can make accurate knowledge recommendations based on the government knowledge system and provide editing and updating capabilities. Throughout the process, it aims at knowledge distribution, knowledge updating, version management, knowledge evaluation, and the real-time and accuracy of the recommendation system.

[0113] The knowledge review and sharing module 830 can review the collected and fed back knowledge and share knowledge. For example, an effective collaboration mechanism between artificial intelligence and human experts is established. Different parts of government affairs knowledge are reviewed by experts in the corresponding parts. After passing the review, they will be shared as government affairs knowledge, thereby ensuring the authority and final quality of knowledge.

[0114] The above-mentioned government knowledge governance system can improve the quality and efficiency of government services through efficient knowledge management and intelligent technology.

[0115] In other embodiments, a computer-readable storage medium stores computer program instructions, which implement the steps of the methods in the above-mentioned embodiments when executed by a processor. It should be understood by those skilled in the art that the embodiments of the present disclosure may be provided as methods, devices, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable non-transient storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0116] In some embodiments, a computer program product is protected, including a computer program or an instruction, which implements the above method when the computer program or instruction is executed by a processor. The computer program product includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by the CPU, the above functions defined in the method of the embodiment of the present disclosure are executed.

[0117] The present disclosure is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present disclosure. It should be understood that each process and / or block in the flowchart and / or block diagram and the combination of processes and / or blocks in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0118] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0119] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0120] So far, the present disclosure has been described in detail. In order to avoid obscuring the concept of the present disclosure, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed here.

[0121] The method and apparatus of the present disclosure may be implemented in many ways. For example, the method and apparatus of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above, unless otherwise specifically stated. In addition, in some embodiments, the present disclosure may also be implemented as a program recorded in a recording medium, which includes machine-readable instructions for implementing the method according to the present disclosure. Therefore, the present disclosure also covers a recording medium storing a program for executing the method according to the present disclosure.

[0122] Although some specific embodiments of the present disclosure have been described in detail by way of example, it should be understood by those skilled in the art that the above examples are for illustration only and are not intended to limit the scope of the present disclosure. It should be understood by those skilled in the art that the above embodiments may be modified without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A knowledge processing method, comprising: Build a user knowledge graph based on each user's historical operations on the knowledge in the public knowledge base; Processing the user knowledge graph to obtain a user feature vector and a knowledge feature vector; Based on the user feature vector and the knowledge feature vector, obtaining an embedding vector matrix of multiple candidate knowledge; as well as Based on the target user's historical operations on knowledge and the embedding vector matrix of the plurality of candidate knowledge, knowledge is recommended to the target user.

2. The knowledge processing method according to claim 1, further comprising: In response to at least one of an editing operation and an updating operation performed by the target user on the recommended knowledge, the public knowledge base is updated.

3. The knowledge processing method according to claim 1, wherein: The recommending knowledge to the target user based on the target user's historical operations on knowledge and the embedding vector matrix of the plurality of candidate knowledge includes: Performing information enhancement on the target user's historical operations on knowledge to obtain a first embedding vector matrix corresponding to the target user; Performing self-attention processing on the first embedding vector matrix to obtain a second embedding vector matrix; Using the second embedding vector matrix and the embedding vector matrices of the plurality of candidate knowledge, scoring each of the plurality of candidate knowledge; and Based on the score of each candidate knowledge, knowledge is recommended to the target user.

4. The knowledge processing method according to claim 3, wherein: The information enhancement of the historical operation of the target user on the knowledge to obtain the first embedding vector matrix corresponding to the target user includes: Averaging multiple contexts of each session in the target user's historical operations on knowledge to obtain a first fused embedding vector representation; Performing time-varying information fusion on each session in the target user's historical operation on the knowledge to obtain a second fused embedding vector representation; Performing weighted calculation on the first fused embedding vector representation and the second fused embedding vector representation to obtain a third fused embedding vector representation corresponding to each session; concatenating the third fused embedding vector representation corresponding to each session and the time of each session, and processing them using a feedforward neural network to obtain an embedding vector matrix corresponding to each session; and According to the embedding vector matrix corresponding to each session, a first embedding vector matrix corresponding to the target user is obtained.

5. The knowledge processing method according to claim 3, wherein: The performing self-attention processing on the first embedding vector matrix to obtain a second embedding vector matrix comprises: The first embedding vector matrix is ​​normalized and indexed by using the query parameter matrix, key parameter matrix, and value parameter matrix of the self-attention algorithm to obtain the second embedding vector matrix.

6. The knowledge processing method according to claim 1, wherein: The processing of the user knowledge graph to obtain a user feature vector and a knowledge feature vector comprises: The user knowledge graph is processed using a bipartite graph to obtain the user feature vector and the knowledge feature vector.

7. The knowledge processing method according to claim 2, wherein: Updating the public knowledge base further includes: In response to the recommended knowledge being reviewed and approved, public knowledge is formed; and The public knowledge base is updated using the public knowledge.

8. The knowledge processing method according to claim 7, wherein: The knowledge in response to the recommendation is reviewed and approved, and the public knowledge is formed, including: classifying the recommended knowledge; and The recommended knowledge is reviewed and approved according to its category to form the public knowledge.

9. The knowledge processing method according to any one of claims 1 to 8, further comprising: Distributed collection of original knowledge; Performing data conversion and preprocessing on the original knowledge; Extract the original knowledge after data conversion and preprocessing around the business domain to form a structured knowledge representation; as well as The structured knowledge representation is sorted to obtain the public knowledge base.

10. A knowledge processing system, comprising a knowledge recommendation module, wherein the knowledge recommendation module comprises: A construction unit, configured to construct a user knowledge graph according to each user's historical operations on the knowledge in the public knowledge base; A processing unit, configured to process the user knowledge graph to obtain a user feature vector and a knowledge feature vector; A candidate unit, configured to obtain an embedding vector matrix of a plurality of candidate knowledge based on the user feature vector and the knowledge feature vector; as well as The recommendation unit is configured to recommend knowledge to the target user based on the target user's historical operations on knowledge by using the user feature vector and the knowledge feature vector.

11. The knowledge processing system according to claim 10, further comprising: The sharing module is configured to update the public knowledge base in response to at least one of an editing operation and an updating operation performed by the target user on the recommended knowledge.

12. The knowledge processing system according to claim 10 or 11, further comprising: The acquisition module is configured to collect original knowledge in a distributed manner, perform data conversion and preprocessing on the original knowledge, extract knowledge from the original knowledge that has undergone data conversion and preprocessing around the business domain to form a structured knowledge representation, organize the structured knowledge representation, and obtain the public knowledge base.

13. A knowledge processing system, comprising: Memory; as well as A processor coupled to the memory, wherein the processor is configured to execute the knowledge processing method according to any one of claims 1 to 9 based on instructions stored in the memory.

14. A computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the knowledge processing method as claimed in any one of claims 1 to 9.

15. A computer program product, comprising a computer program or instructions, wherein when the computer program or instructions are executed by a processor, the knowledge processing method according to any one of claims 1 to 9 is implemented.

Images