Knowledge management method and system based on AI intelligent recommendation

Abstract

The invention relates to a knowledge management method and system based on AI intelligent recommendation, and belongs to the technical field of artificial intelligence and knowledge management. The method comprises the steps of obtaining file data of a user side and preprocessing the file data through a preprocessing model to obtain preprocessed data; according to the preprocessed data, feature data and relational data are obtained through calculation of a joint extraction model, and a knowledge graph is constructed; creating metadata and an identifier, performing data association to obtain associated data, and storing the associated data into a graph structure database; obtaining search semantic data, calculating correlation among the knowledge maps, and performing database search according to the correlation to obtain a search result; according to the retrieval semantic data, the graph structure database is retrieved through the intelligent recommendation model to obtain recommendation data, and the recommendation data is transmitted to a user side recommendation interface. And intelligent knowledge recommendation and knowledge management of the enterprise knowledge management system are realized.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence and knowledge management technology, and specifically relates to a knowledge management method and system based on AI intelligent recommendation. Background Art

[0002] With the rapid development of information technology, especially the continuous advancement of big data and artificial intelligence technology, the field of knowledge management is undergoing a revolution. In today's information and digital age, knowledge management has become the key to improving the competitiveness of enterprises and organizations. With the rapid development of artificial intelligence technology, the application of AI-based intelligent recommendation systems in the field of knowledge management is becoming more and more extensive.

[0003] Current problems:

[0004] (1) The knowledge content in the current enterprise knowledge management system is complex, the amount of information is too large, and the information is updated quickly. The traditional database storage method will affect the efficiency of users in finding the required knowledge objects, thereby affecting the transmission and sharing of knowledge.

[0005] (2) The functions of current enterprise knowledge management systems are relatively simple. The user end only compares and searches the database based on the search semantic data entered by the user, and returns the search results to the human-computer interaction interface. It does not implement intelligent recommendations based on the user's search semantics, transmit related file data to the human-computer interaction interface, and provide users with upstream and downstream information to form a complete knowledge system. Summary of the invention

[0006] In order to solve the above problems existing in the prior art, the present invention provides a knowledge management method and system based on AI intelligent recommendation.

[0007] The purpose of the present invention can be achieved through the following technical solutions:

[0008] S1: acquiring file data of the user end through the internal interface of the system, storing the file data in a buffer pool, and sequentially transferring the file data in the buffer pool to a preprocessing model for preprocessing to obtain preprocessed data;

[0009] S2: obtaining feature data and relationship data by calculating the joint extraction model according to the preprocessed data, and constructing a knowledge graph according to the relationship data and the feature data;

[0010] S3: creating metadata and an identifier according to the file data, associating the identifier with the metadata to obtain identification metadata, then associating the identification metadata with feature data in the knowledge graph to obtain associated data, and storing the knowledge graph, the file data and the identification metadata into a graph structure database according to the associated data;

[0011] S4: Acquire the search semantic data of the user end, extract the knowledge graph through the joint extraction model according to the search semantic data to obtain a conditional knowledge graph, calculate the graph structure database through a search algorithm according to the conditional knowledge graph to obtain search results, and transmit the search results to the user end search page;

[0012] S5: According to the retrieval semantic data, the graph structure database is searched through the intelligent recommendation model to obtain recommendation data, and the recommendation data is transmitted to the user-side recommendation page through the system internal interface.

[0013] Specifically, the preprocessing model is used to unify the file data in different formats, and the specific steps are:

[0014] S201: first read the first four bytes of the file data and convert them into hexadecimal numbers, then compare the hexadecimal numbers with the file type comparison table, and output the file type identification result if the comparison is successful, otherwise read the file name character string of the file data;

[0015] S202: splitting the file name string according to the separator to obtain a file suffix name, comparing the file suffix name with a suffix name file type comparison table to obtain the file type identification result, and labeling the file data according to the file type identification result to obtain the labeled file data;

[0016] S203: Obtain a code recognition result by calculating the target file data through a code recognition module, and convert the code format of the file data through a conversion function according to the code recognition result to obtain converted file data;

[0017] S204: Performing text extraction by a text extraction unit according to the conversion file data to obtain the preprocessed data.

[0018] Specifically, the joint extraction model is composed of an encoder and a decoder. The feature data and the relationship data are extracted in a linked manner by using an end-to-end single model. The preprocessed data and the preset relationship data are first concatenated to obtain a concatenated sequence, and the concatenated sequence is encoded by the encoder to obtain an output sequence. The output sequence is then decoded by a relationship decoder to obtain the relationship data, and the output sequence is decoded by an entity recognition decoder to obtain the feature data. The specific operation expression of the relationship decoder is:

[0019] O r =ReLu(h*w+b)

[0020] Where Or is the relational data, h is the final hidden layer vector, w is the linear mapping matrix, and b is the bias vector.

[0021] Specifically, the specific operation expression of the entity recognition decoder is:

[0022] O n =ReLu(H s *W N +B N )

[0023] Among them, O n is the characteristic data, H s is the embedding representation of the corresponding file data word sequence output by the final hidden layer, W N is the linear mapping matrix, B N is the bias matrix.

[0024] Specifically, the search algorithm retrieves the graph structure database based on the knowledge graph to obtain the search results, calculates the correlation between the knowledge graph of each file data and the conditional knowledge graph, and forms an array for each correlation, obtains the maximum value of the array through binary search, finds the knowledge graph that matches the maximum value based on the maximum value, and searches the graph structure database based on the feature data of the knowledge graph to find the file data corresponding to the feature data, and the file data is the search result.

[0025] Specifically, the correlation is calculated based on the standard accuracy and recall rate, and the correlation calculation formula is:

[0026]

[0027] Where Re is the relevance, b is an important hyperparameter for balancing precision and recall, N and M represent the number of fact knowledge carrying paths detected in the knowledge graph and the conditional knowledge graph, i and j are variable parameters, and Q(P i , P j ) is a function used to calculate the similarity of a pair of knowledge carrying paths, P i is the feature data knowledge carrying path, P j Carrying paths for relational data knowledge.

[0028] Specifically, the intelligent recommendation model calculation process is:

[0029] S701: Calculate the correlation between the knowledge graph of each file data and the knowledge graph of the search semantic data, and store each correlation in an array;

[0030] S702: Sorting the array by a heap sorting algorithm to obtain an ordered array;

[0031] S703: Compare the correlations in the ordered array with preset thresholds in sequence to obtain matching correlations, and retrieve the recommended data from the graph structure database according to the matching correlations.

[0032] A knowledge management system based on AI intelligent recommendation, including: a preprocessing module, a building module, a matching module, a search module, and a recommendation module;

[0033] The preprocessing module is used to obtain file data of the user end through the system internal interface, store the file data in the buffer pool, and sequentially transfer the file data in the buffer pool to the preprocessing model for preprocessing to obtain preprocessed data;

[0034] The construction module is used to calculate feature data and relationship data according to the preprocessed data through a joint extraction model, and construct a knowledge graph according to the relationship data and the feature data;

[0035] The matching module is used to create metadata and an identifier according to the file data, associate the identifier with the metadata to obtain identification metadata, and then associate the identification metadata with feature data in the knowledge graph to obtain associated data, and store the knowledge graph, the file data and the identification metadata into a graph structure database according to the associated data;

[0036] The search module is used to obtain the search semantic data of the user end, extract the knowledge graph through the joint extraction model according to the search semantic data to obtain the conditional knowledge graph, calculate the graph structure database through the search algorithm according to the conditional knowledge graph to obtain the search results, and transmit the search results to the user end search page;

[0037] The recommendation module is used to retrieve the graph structure database through an intelligent recommendation model according to the retrieval semantic data to obtain recommendation data, and transmit the recommendation data to the user-side recommendation page through the system internal interface.

[0038] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned knowledge management method based on AI intelligent recommendation when executing the program.

[0039] A storage medium containing computer executable instructions, characterized in that the computer executable instructions are used to execute the knowledge management method based on AI intelligent recommendation as described above when executed by a computer processor.

[0040] The beneficial effects of the present invention are:

[0041] 1. Extract the knowledge graph of dynamically acquired file data and search semantic data, calculate the correlation between the knowledge graph of dynamically acquired file data and the knowledge graph of search semantic data, and search the database based on the correlation, which greatly improves the retrieval accuracy and retrieval speed.

[0042] 2. By calculating the correlation between the knowledge graph of file data in the database and the knowledge graph of search semantic data, by comparing the correlation with the preset threshold and outputting the file data corresponding to the correlation greater than or equal to the preset threshold, and displaying it in order from large to small on the human-computer interaction interface, intelligent recommendation is achieved, providing users with upstream and downstream information integration, and forming a complete information system.

[0043] 3. The acquired file data is stored in a graph structure database, which can efficiently represent and query highly connected data, optimize traversal operations, process data updates and queries in real time, and improve the responsiveness of the application. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0045] Figure 1 The figure is a flow chart of a knowledge management method based on AI intelligent recommendation of the present invention. DETAILED DESCRIPTION

[0046] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation methods, structures, features and effects of the present invention are described in detail below in conjunction with the accompanying drawings and preferred embodiments.

[0047] See also Figure 1 , a knowledge management method based on AI intelligent recommendation,

[0048] S1: acquiring file data of the user end through the internal interface of the system, storing the file data in a buffer pool, and sequentially transferring the file data in the buffer pool to a preprocessing model for preprocessing to obtain preprocessed data;

[0049] S2: obtaining feature data and relationship data by calculating the joint extraction model according to the preprocessed data, and constructing a knowledge graph according to the relationship data and the feature data;

[0050] S3: creating metadata and an identifier according to the file data, associating the identifier with the metadata to obtain identification metadata, then associating the identification metadata with feature data in the knowledge graph to obtain associated data, and storing the knowledge graph, the file data and the identification metadata into a graph structure database according to the associated data;

[0051] S4: Acquire the search semantic data of the user end, extract the knowledge graph through the joint extraction model according to the search semantic data to obtain a conditional knowledge graph, calculate the graph structure database through a search algorithm according to the conditional knowledge graph to obtain search results, and transmit the search results to the user end search page;

[0052] S5: According to the retrieval semantic data, the graph structure database is searched through the intelligent recommendation model to obtain recommendation data, and the recommendation data is transmitted to the user-side recommendation page through the system internal interface.

[0053] Specifically, the preprocessing model is used to unify the file data in different formats, and the specific steps are:

[0054] S201: first read the first four bytes of the file data and convert them into hexadecimal numbers, then compare the hexadecimal numbers with the file type comparison table, and output the file type identification result if the comparison is successful, otherwise read the file name character string of the file data;

[0055] S202: splitting the file name string according to the separator to obtain a file suffix name, comparing the file suffix name with a suffix name file type comparison table to obtain the file type identification result, and labeling the file data according to the file type identification result to obtain the labeled file data;

[0056] S203: Obtain a code recognition result by calculating the target file data through a code recognition module, and convert the code format of the file data through a conversion function according to the code recognition result to obtain converted file data;

[0057] S204: Performing text extraction by a text extraction unit according to the conversion file data to obtain the preprocessed data.

[0058] In this embodiment, the encoding recognition module is mainly used to identify the encoding of file data. Since the different intelligent terminal systems used by users lead to differences in the incoming file encodings, different files need to be unified before constructing the knowledge graph of file data; the conversion function uses an icon function provided by the libiconv library to support mutual conversion between numerous character encodings.

[0059] Specifically, the joint extraction model is composed of an encoder and a decoder. The feature data and the relationship data are extracted in a linked manner by using an end-to-end single model. The preprocessed data and the preset relationship data are first concatenated to obtain a concatenated sequence, and the concatenated sequence is encoded by the encoder to obtain an output sequence. The output sequence is then decoded by a relationship decoder to obtain the relationship data, and the output sequence is decoded by an entity recognition decoder to obtain the feature data. The specific operation expression of the relationship decoder is:

[0060] O r =ReLu(h*w+b)

[0061] Where Or is the relational data, h is the final hidden layer vector, w is the linear mapping matrix, and b is the bias vector.

[0062] Specifically, the specific operation expression of the entity recognition decoder is:

[0063] O n =ReLu(H s *W N +B N )

[0064] Among them, O n is the characteristic data, H s is the embedding representation of the corresponding file data word sequence output by the final hidden layer, W N is the linear mapping matrix, B N is the bias matrix.

[0065] In this embodiment, the encoding recognition module is a pre-trained model based on a multi-layer multi-head self-attention encoder.

[0066] Specifically, the search algorithm retrieves the graph structure database based on the knowledge graph to obtain the search results, calculates the correlation between the knowledge graph of each file data and the conditional knowledge graph, and forms an array for each correlation, obtains the maximum value of the array through binary search, finds the knowledge graph that matches the maximum value based on the maximum value, and searches the graph structure database based on the feature data of the knowledge graph to find the file data corresponding to the feature data, and the file data is the search result.

[0067] In this embodiment, the knowledge graph of dynamically acquired file data and search semantic data is extracted, and the correlation between the knowledge graph of dynamically acquired file data and the knowledge graph of search semantic data is calculated, and the database is searched based on the correlation, which greatly improves the retrieval accuracy and retrieval speed.

[0068] Specifically, the correlation is calculated based on the standard accuracy and recall rate, and the correlation calculation formula is:

[0069]

[0070] Where Re is the relevance, b is an important hyperparameter for balancing precision and recall, N and M represent the number of fact knowledge carrying paths detected in the knowledge graph and the conditional knowledge graph, i and j are variable parameters, and Q(P i , P j ) is a function used to calculate the similarity of a pair of knowledge carrying paths, P i is the feature data knowledge carrying path, P j Carrying paths for relational data knowledge.

[0071] In this embodiment, by calculating the correlation between the knowledge graph of the file data in the database and the knowledge graph of the search semantic data, by comparing the correlation with a preset threshold and outputting the file data corresponding to the correlation greater than or equal to the preset threshold, and displaying them in order from large to small on the human-computer interaction interface, intelligent recommendation is achieved, providing users with upstream and downstream information integration, and forming a complete information system.

[0072] Specifically, the intelligent recommendation model calculation process is:

[0073] S701: Calculate the correlation between the knowledge graph of each file data and the knowledge graph of the search semantic data, and store each correlation in an array;

[0074] S702: Sorting the array by a heap sorting algorithm to obtain an ordered array;

[0075] S703: Compare the correlations in the ordered array with preset thresholds in sequence to obtain matching correlations, and retrieve the recommended data from the graph structure database according to the matching correlations.

[0076] In this embodiment, the acquired file data is stored in a graph structure database, which can efficiently represent and query highly connected data, optimize traversal operations, process data updates and queries in real time, and improve application responsiveness.

[0077] A knowledge management system based on AI intelligent recommendation, including: a preprocessing module, a building module, a matching module, a search module, and a recommendation module;

[0078] The preprocessing module is used to obtain file data of the user end through the system internal interface, store the file data in the buffer pool, and sequentially transfer the file data in the buffer pool to the preprocessing model for preprocessing to obtain preprocessed data;

[0079] The construction module is used to calculate feature data and relationship data according to the preprocessed data through a joint extraction model, and construct a knowledge graph according to the relationship data and the feature data;

[0080] The matching module is used to create metadata and an identifier according to the file data, associate the identifier with the metadata to obtain identification metadata, and then associate the identification metadata with feature data in the knowledge graph to obtain associated data, and store the knowledge graph, the file data and the identification metadata into a graph structure database according to the associated data;

[0081] The search module is used to obtain the search semantic data of the user end, extract the knowledge graph through the joint extraction model according to the search semantic data to obtain the conditional knowledge graph, calculate the graph structure database through the search algorithm according to the conditional knowledge graph to obtain the search results, and transmit the search results to the user end search page;

[0082] The recommendation module is used to retrieve the graph structure database through an intelligent recommendation model according to the retrieval semantic data to obtain recommendation data, and transmit the recommendation data to the user-side recommendation page through the system internal interface.

[0083] In this embodiment, the metadata includes file title, author, creation date, modification date, type, format, and language.

[0084] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned knowledge management method based on AI intelligent recommendation when executing the program.

[0085] A storage medium containing computer executable instructions, characterized in that the computer executable instructions are used to execute the knowledge management method based on AI intelligent recommendation as described above when executed by a computer processor. The program code contained in the method in the embodiment of the present invention can be transmitted by any appropriate medium, including but not limited to wireless, wires, optical cables, etc., or any suitable combination of the above. The computer program code for performing the operation of the present invention can be written in one or more programming languages ​​or a combination thereof, and the programming language includes an object-oriented programming language-such as Java, Smalltalk, C++, and also includes a conventional procedural programming language-such as "C" language or similar programming language. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (for example, using an Internet service provider to connect through the Internet).

[0086] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A knowledge management method based on AI intelligent recommendation, characterized in that: The following steps are involved: S1: acquiring file data of the user end through the internal interface of the system, storing the file data in a buffer pool, and sequentially transferring the file data in the buffer pool to a preprocessing model for preprocessing to obtain preprocessed data; S2: Calculating feature data and relationship data through a joint extraction model according to the preprocessed data, and constructing a knowledge graph according to the relationship data and the feature data; S3: creating metadata and an identifier according to the file data, associating the identifier with the metadata to obtain identification metadata, then associating the identification metadata with feature data in the knowledge graph to obtain associated data, and storing the knowledge graph, the file data and the identification metadata into a graph structure database according to the associated data; S4: Acquire the search semantic data of the user end, extract the knowledge graph through the joint extraction model according to the search semantic data to obtain a conditional knowledge graph, calculate the graph structure database through a search algorithm according to the conditional knowledge graph to obtain search results, and transmit the search results to the user end search page; S5: According to the retrieval semantic data, the graph structure database is searched through the intelligent recommendation model to obtain recommendation data, and the recommendation data is transmitted to the user-side recommendation page through the system internal interface.

2. The knowledge management method based on AI intelligent recommendation according to claim 1 is characterized in that: The preprocessing model is used to unify the file data in different formats, and the specific steps are: S201: first read the first four bytes of the file data and convert them into hexadecimal numbers, then compare the hexadecimal numbers with the file type comparison table, and output the file type identification result if the comparison is successful, otherwise read the file name character string of the file data; S202: splitting the file name string according to the separator to obtain a file suffix name, comparing the file suffix name with a suffix name file type comparison table to obtain the file type identification result, and labeling the file data according to the file type identification result to obtain the labeled file data; S203: Obtain a code recognition result by calculating the target file data through a code recognition module, and convert the code format of the file data through a conversion function according to the code recognition result to obtain converted file data; S204: Performing text extraction by a text extraction unit according to the conversion file data to obtain the preprocessed data.

3. The knowledge management method based on AI intelligent recommendation according to claim 1 is characterized in that: The joint extraction model is composed of an encoder and a decoder. The feature data and the relationship data are extracted in a linked manner by using an end-to-end single model. The preprocessed data and the preset relationship data are first concatenated to obtain a concatenated sequence, and the concatenated sequence is encoded by the encoder to obtain an output sequence. The output sequence is then decoded by a relationship decoder to obtain the relationship data, and the output sequence is then decoded by an entity recognition decoder to obtain the feature data. The specific operation expression of the relationship decoder is: O r =ReLu(h*w+b) Among them, r is the relational data, h is the final hidden layer vector, w is the linear mapping matrix, and b is the bias vector.

4. The knowledge management method based on AI intelligent recommendation according to claim 3 is characterized in that: The specific operation expression of the entity recognition decoder is: O n =ReLu(Hs*W N +B N ) Among them, O n is the characteristic data, H s is the embedding representation of the corresponding file data word sequence output by the final hidden layer, W N is the linear mapping matrix, B N is the bias matrix.

5. The knowledge management method based on AI intelligent recommendation according to claim 1 is characterized in that: The search algorithm retrieves the graph structure database based on the knowledge graph to obtain the search results, calculates the correlation between the knowledge graph of each file data and the conditional knowledge graph, and forms an array with each correlation, obtains the maximum value of the array through binary search, finds the knowledge graph that matches it based on the maximum value, and searches the graph structure database based on the feature data of the knowledge graph to find the file data corresponding to the feature data, and the file data is the search result.

6. The knowledge management method based on AI intelligent recommendation according to claim 5 is characterized in that: The correlation is calculated based on the standard accuracy and recall rate, and the correlation calculation formula is: Where Re is the relevance, b is an important hyperparameter for balancing precision and recall, N and M represent the number of fact knowledge carrying paths detected in the knowledge graph and the conditional knowledge graph, i and j are variable parameters, and Q(P i ,P j ) is a function used to calculate the similarity of a pair of knowledge carrying paths, P i is the feature data knowledge carrying path, P j Carrying paths for relational data knowledge.

7. The knowledge management method based on AI intelligent recommendation according to claim 1, characterized in that: The calculation process of the intelligent recommendation model is as follows: S701: Calculate the correlation between the knowledge graph of each file data and the knowledge graph of the search semantic data, and store each correlation in an array; S702: Sorting the array by a heap sorting algorithm to obtain an ordered array; S703: Compare the correlations in the ordered array with preset thresholds in sequence to obtain matching correlations, and retrieve the recommended data from the graph structure database according to the matching correlations.

8. A knowledge management system based on AI intelligent recommendation, characterized in that: include: Preprocessing module, building module, matching module, search module, recommendation module; The preprocessing module is used to obtain file data of the user end through the system internal interface, store the file data in the buffer pool, and sequentially transfer the file data in the buffer pool to the preprocessing model for preprocessing to obtain preprocessed data; The construction module is used to calculate feature data and relationship data according to the preprocessed data through a joint extraction model, and construct a knowledge graph according to the relationship data and the feature data; The matching module is used to create metadata and an identifier according to the file data, associate the identifier with the metadata to obtain identification metadata, and then associate the identification metadata with feature data in the knowledge graph to obtain associated data, and store the knowledge graph, the file data and the identification metadata into a graph structure database according to the associated data; The search module is used to obtain the search semantic data of the user end, extract the knowledge graph through the joint extraction model according to the search semantic data to obtain the conditional knowledge graph, calculate the graph structure database through the search algorithm according to the conditional knowledge graph to obtain the search results, and transmit the search results to the user end search page; The recommendation module is used to retrieve the graph structure database through an intelligent recommendation model according to the retrieval semantic data to obtain recommendation data, and transmit the recommendation data to the user-side recommendation page through the system internal interface.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the knowledge management method based on AI intelligent recommendation as described in any one of claims 1 to 7 is implemented.

10. A storage medium containing computer executable instructions, characterized in that: The computer executable instructions, when executed by a computer processor, are used to execute the knowledge management method based on AI intelligent recommendation as described in any one of claims 1 to 7.

Images