Query enhancement technology-based operation and maintenance knowledge base question answering system and method

By using an operations and maintenance knowledge base question-and-answer system based on query enhancement technology, user input questions are expanded and vectorized, which solves the problem of inaccurate answers in the operations and maintenance knowledge base, improves the efficiency and accuracy of operations and maintenance management, and discovers more solutions.

CN119046435APending Publication Date: 2024-11-29SUNING CONSUMER FINANCE CO LTD
View PDF 4 Cites 0 Cited by

Patent Information

Application Number
CN202411217350.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-02
Publication Date
2024-11-29

AI Technical Summary

Technical Problem

The existing operations and maintenance knowledge base has a low accuracy rate when answering user queries, mainly because the user's query statement is not accurate enough or the knowledge base's query algorithm is not intelligent enough to accurately understand the user's needs and return the most relevant answer.

Method used

The operation and maintenance knowledge base question answering system adopts query enhancement technology, including a query enhancement module, a feature extraction module, a vector database module, and a fine ranking module. It expands and enhances user input questions through strategies such as synonym expansion, related word recommendation, and contextual information fusion, and uses N-gram analysis and embedding model for vectorization processing. It also combines the existing knowledge base for retrieval and splicing to improve the accuracy of the answers.

Benefits of technology

It improves the accuracy of the answers output by the operation and maintenance knowledge base question and answer system, enhances the efficiency of operation and maintenance management, enables a better understanding of user needs, discovers hidden knowledge and solutions, broadens knowledge horizons, and adapts to the latest information.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119046435A_ABST
    Figure CN119046435A_ABST
Patent Text Reader

Abstract

The invention discloses an operation and maintenance knowledge base question answering system and method based on a Query enhancement technology. The system comprises a vector database module used for storing uploaded operation and maintenance document fragments; the Query enhancement module is used for expanding and enhancing keywords in the questions input by the user according to the operation and maintenance document fragments in the vector database module to obtain more operation and maintenance document fragments related to the questions input by the user; the feature extraction module is used for performing word segmentation on all operation and maintenance document fragments related to questions input by the user; and the fine ranking module is used for matching the segmented words with the operation and maintenance document fragments stored in the vector database module according to a ranking algorithm, ranking the matched operation and maintenance document fragments according to a high-low sequence of matching degrees, taking the matched operation and maintenance document fragments as answers of the question input by the user, and feeding back the answers to the user. According to the method, the answer output accuracy of the operation and maintenance knowledge base question-answering system is improved, so that the operation and maintenance management efficiency is effectively improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of operation and maintenance knowledge question and answer, and in particular to an operation and maintenance knowledge base question and answer system and method based on query enhancement technology. Background Art

[0002] In operation and maintenance work, operation and maintenance personnel often need to query solutions from the knowledge base to deal with various problems. However, the existing operation and maintenance knowledge base often has the problem of low accuracy when answering user queries. This is mainly because the query statement entered by the user may not be accurate and comprehensive enough, or the query algorithm of the knowledge base is not smart enough to accurately understand the user's needs and return the most relevant answer. Summary of the invention

[0003] In view of the problems existing in the prior art, the present invention provides an operation and maintenance knowledge base question and answer system and method based on Query enhancement technology, which improves the accuracy of answers output by the operation and maintenance knowledge base question and answer system, thereby effectively improving the efficiency of operation and maintenance management.

[0004] To achieve the above technical objectives, the present invention adopts the following technical solutions: an operation and maintenance knowledge base question-answering system based on query enhancement technology, comprising: a query enhancement module, a feature extraction module, a vector database module, and a precision sorting module;

[0005] The vector database module is used to update and store uploaded operation and maintenance document fragments in real time;

[0006] The query enhancement module expands and enhances the keywords in the user input question according to the operation and maintenance document fragments in the vector database module to obtain more operation and maintenance document fragments related to the user input question;

[0007] The feature extraction module is used to segment all operation and maintenance document fragments related to the user input question into words;

[0008] The refined sorting module is used to match the segmented words with the operation and maintenance document fragments stored in the vector database module according to the sorting algorithm, and sort the matched operation and maintenance document fragments in order of the degree of matching, and feed back to the user as the answer to the user input question.

[0009] Furthermore, the strategies used by the query enhancement module to expand and enhance the questions input by the user include: synonym expansion strategy, related word recommendation strategy, and context information fusion strategy;

[0010] The synonym expansion strategy is used to find synonyms of keywords in the user input question and add them to the user input question;

[0011] The related word recommendation strategy recommends words related to the keywords based on the association between the keywords in the user input question and the operation and maintenance document fragments in the vector database module, and adds them to the user input question;

[0012] The context information fusion strategy adds relevant information to the user input question by considering the context of the operational document fragments in the user query vector database module.

[0013] Furthermore, the feature extraction module uses an N-gram analysis method to segment all operation and maintenance document fragments related to the user input question into a number of word combinations.

[0014] Furthermore, it also includes: a data processing structured segmentation module; the data processing structured segmentation module divides various operation and maintenance documents uploaded by users according to the title level, and retains the context and structural information of the text corresponding to the title. At the same time, the overall context of the operation and maintenance document and the relationship between sentences and phrases within the text are retained through the operation and maintenance document structured decomposition algorithm.

[0015] Furthermore, after the operation and maintenance document fragments segmented by the data processing structured segmentation module are subjected to feature extraction by the feature extraction module, the segmented operation and maintenance document fragments are vectorized using the embedding model and stored in the vector database module.

[0016] Furthermore, it also includes: a stock knowledge base module and a corpus splicing module;

[0017] The stock knowledge base module searches the stock knowledge base for the user input question to obtain the operation and maintenance document content related to the user input question;

[0018] The corpus splicing module is used to perform prompt splicing on the operation and maintenance document content related to the user input question obtained from the stock knowledge base module and the answer to the user input question in the refined sorting module, update the answer to the user input question, and feed it back to the user.

[0019] Furthermore, the stock knowledge base module searches the stock knowledge base for the user input question, and the specific process of obtaining the operation and maintenance document content related to the user input question is: use the NER model to perform entity query on the user input question, extract the keywords of the user input question, classify the keywords according to semantic categories, and convert the keywords into structured data, call the retrieval interface of the stock knowledge base module, and match the operation and maintenance document content related to the user input question from the stock knowledge base.

[0020] Furthermore, the request format of the user input question is in XML format or JSON format, and supports jsonp cross-domain access.

[0021] Furthermore, the present invention also provides a question-answering method of the operation and maintenance knowledge base question-answering system based on the query enhancement technology, which specifically comprises the following steps:

[0022] Step S1: The operation and maintenance documents uploaded by the user are segmented by the data processing structured segmentation module, the segmented operation and maintenance document fragments are segmented by the feature extraction module, and the segmented operation and maintenance document fragments are vectorized using the embedding model and stored in the vector database module;

[0023] Step S2: The query enhancement module expands and enhances the user input question to obtain more operation and maintenance document fragments related to the user input question;

[0024] Step S3: the feature extraction module is used to perform word segmentation on all operation and maintenance document fragments related to the user input question;

[0025] Step S4: The sorting module matches the segmented words with the operation and maintenance document fragments stored in the vector database module, and sorts the matched operation and maintenance document fragments in order of matching degree, and feeds back to the user as the answer to the user input question.

[0026] Furthermore, if there is an existing knowledge base module, the NER model is used to perform entity query on the user input question, extract the keywords of the user input question, classify the keywords according to semantic categories, and convert the keywords into structured data, call the retrieval interface of the existing knowledge base module, match the operation and maintenance document content related to the user input question from the existing knowledge base module, and use the corpus splicing module to perform prompt splicing on the operation and maintenance document content related to the user input question obtained from the existing knowledge base and the answer to the user input question output by the refinement module, update the answer to the user input question, and feedback to the user.

[0027] Compared with the prior art, the present invention has the following beneficial effects: the operation and maintenance knowledge base question and answer system and method based on query enhancement technology of the present invention analyzes and enhances the user input questions through the query enhancement module, automatically supplements the context information of the user input questions and the synonyms of the keywords in the user input questions, so that the answers of the operation and maintenance knowledge base question and answer system can more accurately match the user input questions, better understand the user's real needs, and thus improve the accuracy of the questions and answers; at the same time, the query enhancement module can also mine hidden knowledge and related information related to the keywords in the user input questions from the operation and maintenance document fragments in the vector database module through semantic analysis and association rules, helping users to discover some problem solutions or lessons learned that may not have been noticed before, thereby broadening the knowledge horizon and improving the operation and maintenance efficiency; in addition, the vector database module is used to update the uploaded operation and maintenance document fragments in real time, so as to ensure that the vector database module can provide the latest and most effective information, so that when the query enhancement module performs user input question enhancement, it can expand and enhance according to the latest operation and maintenance document fragments in the vector database module, and ensure that the operation and maintenance knowledge base question and answer system has strong adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 It is a schematic diagram of the operation and maintenance knowledge base question and answer system based on the query enhancement technology of the present invention;

[0029] Figure 2 A schematic diagram of segmenting operation and maintenance document fragments by the data processing structured segmentation module in the present invention. DETAILED DESCRIPTION

[0030] The technical solution of the present invention is further explained below in conjunction with the accompanying drawings.

[0031] like Figure 1The present invention provides an operation and maintenance knowledge base question and answer system based on query enhancement technology, including: a query enhancement module, a feature extraction module, a vector database module, and a precision sorting module; the vector database module is used to update and store uploaded operation and maintenance document fragments in real time, so as to ensure that the vector database module can always provide the latest and most effective document fragments, so that the answers to user input questions always remain the optimal solutions; the query enhancement module expands and enhances the keywords in the user input questions according to the operation and maintenance document fragments in the vector database module, and obtains more operation and maintenance document fragments related to the user input questions. The present invention queries the user input questions in the vector database module through the query enhancement module, and adopts efficient indexing and search algorithms to ensure that relevant operation and maintenance document fragments are found quickly and accurately, thereby improving the efficiency of operation and maintenance management; at the same time, the present invention improves the utilization rate and value of knowledge by establishing the association between problems and solutions and the dependency between different operation and maintenance tasks. The request format of the user input question in the present invention is XML format or JSON format, and supports JSONP cross-domain access; the feature extraction module is used to segment all operation and maintenance document fragments related to the user input question into words; the fine sorting module is used to match the segmented words with the operation and maintenance document fragments stored in the vector database module according to the sorting algorithm, specifically, the segmented words and the keywords of the operation and maintenance document fragments stored in the vector database module are calculated similarity according to the sorting algorithm such as the bubble sorting algorithm and the radix sorting algorithm, and if the similarity is higher than the threshold, it is considered that the segmented words match the operation and maintenance document fragments in the vector database module, and the matched operation and maintenance document fragments are sorted in order of the degree of matching, as the answer to the user input question, and fed back to the user. The operation and maintenance knowledge base question and answer system of the present invention improves the accuracy of the output answer of the operation and maintenance knowledge base question and answer system, thereby effectively improving the efficiency of operation and maintenance management.

[0032] The strategies used by the query enhancement module in the present invention to expand and enhance the questions input by the user include: synonym expansion strategy, related word recommendation strategy, and context information fusion strategy; the synonym expansion strategy is used to find synonyms of keywords in the user input question and add them to the user input question to expand the search scope. For example, if the user input question contains "server failure", synonyms such as "server downtime" and "server abnormality" can be added to the input question, so that the answer of the operation and maintenance knowledge base question and answer system can more accurately match the user input question, better understand the user's real needs, and thus improve the accuracy of the question and answer. When the user inputs a vague or incomplete question, the related word recommendation strategy recommends words related to the keywords based on the association between the keywords in the user input question and the operation and maintenance document fragments in the vector database module, and adds them to the user input question to make the query clearer, and then retrieve a more accurate solution. For example, for the user input question of "database connection failure", related words such as "database configuration" and "network problem" can be recommended. The context information fusion strategy considers the context of the operation and maintenance document fragments in the user query vector database module, such as previous query history, current operating environment, etc., and adds relevant information to the user input question to improve accuracy. If the user has previously searched for "server configuration" and then entered "poor server performance", the previous query can be combined to enhance the current user input question to "poor server performance problem related to server configuration". Through semantic analysis and association rules, hidden knowledge and related information related to the keywords in the user input question are mined from the operation and maintenance document fragments in the vector database module, helping users discover some problem solutions or lessons learned that may not have been noticed before, thereby broadening their knowledge horizons and improving operation and maintenance efficiency.

[0033] The feature extraction module in the present invention uses the N-gram analysis method to divide all operation and maintenance document fragments related to the user input questions into several word combinations.

[0034] In addition, the present invention also includes: a data processing structured segmentation module, such as Figure 2The data processing structured segmentation module accurately segments various operation and maintenance documents uploaded by users according to the title level, and retains the context and structural information of the text corresponding to the title, which helps to improve the effect of text vectorization. At the same time, the operation and maintenance document structured decomposition algorithm retains the overall context of the operation and maintenance document and the relationship between sentences and phrases in the text, thereby generating a more comprehensive vector representation and capturing the broader meaning and theme of the text. The structured decomposition process uses models and frameworks designed specifically for relationship extraction tasks, such as RelationBERT, combined with knowledge graphs, external relational databases, and context enhancement methods to ensure that multiple connections between entities are accurately captured in complex texts and understand their dynamic interactions and connections. For example, when the server repairs the patch, not only the "server" and "patch" entities are identified, but also the "repair" relationship between them is extracted. After the operation and maintenance document fragments segmented by the data processing structured segmentation module are extracted by the feature extraction module, the segmented operation and maintenance document fragments are vectorized using the embedding model and stored in the vector database module.

[0035] The present invention also includes: an existing knowledge base module and a corpus splicing module. The existing knowledge base module searches the existing knowledge base for user input questions to obtain operation and maintenance document content related to the user input questions; the corpus splicing module is used to prompt-splice the operation and maintenance document content related to the user input questions obtained from the existing knowledge base module with the answers to the user input questions in the sorting module, update the answers to the user input questions, and feed back to the user.

[0036] The stock knowledge base module searches the stock knowledge base for user input questions and obtains the operation and maintenance document content related to the user input questions in the following specific process: use the NER model to perform entity query on the user input questions, extract the keywords of the user input questions, classify the keywords according to semantic categories, such as operation and maintenance instances, operation and maintenance organizations, operation and maintenance events, operation and maintenance scripts, solutions, etc., and convert the keywords into structured data, call the retrieval interface of the stock knowledge base module, and match the operation and maintenance document content related to the user input questions from the stock knowledge base. The stock knowledge base module is a supplement to the knowledge resources in the vector database. This is because the stock knowledge base module not only accumulates a large amount of knowledge from the operation and maintenance field, but also has a wide range of knowledge coverage and rich and diverse content, which is consistent with the operation and maintenance knowledge system. It can provide more comprehensive information support for the operation and maintenance team, help solve various complex operation and maintenance problems, and make up for the knowledge shortcomings or limitations that may exist in the vector database. In addition, the use of the stock knowledge base module can directly use its existing resources and architecture to output the operation and maintenance document content that matches the user input questions, which can save the cost and time of knowledge base construction.

[0037] The present invention also provides a question-answering method of an operation and maintenance knowledge base question-answering system based on query enhancement technology, which specifically comprises the following steps:

[0038] Step S1: The operation and maintenance documents uploaded by the user include docx, pdf, txt and other types. They are segmented by the data processing structured segmentation module, and the segmented operation and maintenance document fragments are segmented by the feature extraction module. The segmented operation and maintenance document fragments are vectorized using the embedding model and stored in the vector database module. The vector database module is used to update and store the uploaded operation and maintenance document fragments in real time, so as to ensure that the vector database module can provide the latest and most effective information.

[0039] Step S2: The query enhancement module expands and enhances the user input question to obtain more operation and maintenance document fragments related to the user input question. When the query enhancement module performs the user input question enhancement, it can expand and enhance the operation and maintenance document fragments in the latest vector database module to ensure that the operation and maintenance knowledge base question-answering system has strong adaptability.

[0040] Step S3: The feature extraction module is used to perform word segmentation on all operation and maintenance document fragments related to the user input question;

[0041] Step S4: The sorting module matches the segmented words with the operation and maintenance document fragments stored in the vector database module, and sorts the matched operation and maintenance document fragments in order of matching degree, filters out some interference information, and feeds back to the user as the answer to the user input question.

[0042] If there is an existing knowledge base module, use the NER model to perform entity query on the user input question, extract the keywords of the user input question, classify the keywords according to semantic categories, and convert the keywords into structured data, call the retrieval interface of the existing knowledge base module, match the operation and maintenance document content related to the user input question from the existing knowledge base module, and use the corpus splicing module to perform prompt splicing on the operation and maintenance document content related to the user input question obtained from the existing knowledge base and the answer to the user input question output by the refinement module, update the answer to the user input question, and feedback to the user.

[0043] The question-answering method of the operation and maintenance knowledge base question-answering system based on the query enhancement technology of the present invention can not only search according to the keywords of the user input questions, but also deeply explore the potential associations between knowledge, so that the vector database module can provide users with more comprehensive and in-depth knowledge content, helping operation and maintenance personnel to better understand the essence of the problem and provide possible solutions. In addition, the query enhancement module helps to promote the sharing and inheritance of operation and maintenance document fragments in the vector database module. The query enhancement module automatically adds keywords, synonyms, near synonyms, etc. of the input questions through multi-source knowledge, expands the search scope, and uses technologies such as knowledge graphs to explore the potential associations between the query and the knowledge in the vector database module, thereby improving the accuracy of the question and answer; and uses advanced sorting algorithms through the fine sorting module to sort the retrieved question and answer results, and put the most relevant and accurate answers in the front, ultimately improving the accuracy of the user's question and answer.

[0044] In the embodiments disclosed in the present application, the computer storage medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. The computer storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the above. More specific examples of computer storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above.

[0045] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this application can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0046] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should be regarded as the protection scope of the present invention.

Claims

1. An operation and maintenance knowledge base question and answer system based on query enhancement technology, It is characterized in that include: Query enhancement module, feature extraction module, vector database module, and precision sorting module; The vector database module is used to update and store uploaded operation and maintenance document fragments in real time; The query enhancement module expands and enhances the keywords in the user input question according to the operation and maintenance document fragments in the vector database module to obtain more operation and maintenance document fragments related to the user input question; The feature extraction module is used to segment all operation and maintenance document fragments related to the user input question into words; The refined sorting module is used to match the segmented words with the operation and maintenance document fragments stored in the vector database module according to the sorting algorithm, and sort the matched operation and maintenance document fragments in order of the degree of matching, and feed back to the user as the answer to the user input question.

2. According to claim 1, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that The query enhancement module is used to expand and enhance the questions input by the user, including: synonym expansion strategy, related word recommendation strategy, context information fusion strategy; The synonym expansion strategy is used to find synonyms of keywords in the user input question and add them to the user input question; The related word recommendation strategy recommends words related to the keywords based on the association between the keywords in the user input question and the operation and maintenance document fragments in the vector database module, and adds them to the user input question; The context information fusion strategy adds relevant information to the user input question by considering the context of the operational document fragments in the user query vector database module.

3. According to claim 2, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that The feature extraction module uses an N-gram analysis method to segment all operation and maintenance document fragments related to the user input question into a number of word combinations.

4. According to claim 3, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that Also includes: Data processing structured segmentation module: The data processing structured segmentation module divides various operation and maintenance documents uploaded by users according to the title level, and retains the context and structural information of the text corresponding to the title. At the same time, the overall context of the operation and maintenance document and the relationship between sentences and phrases within the text are retained through the operation and maintenance document structured disassembly algorithm.

5. According to claim 4, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that After the operation and maintenance document fragments segmented by the data processing structured segmentation module are subjected to feature extraction by the feature extraction module, the segmented operation and maintenance document fragments are vectorized using the embedding model and stored in the vector database module.

6. According to claim 5, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that Also includes: Stock knowledge base module and corpus splicing module; The stock knowledge base module searches the stock knowledge base for the user input question to obtain the operation and maintenance document content related to the user input question; The corpus splicing module is used to perform prompt splicing on the operation and maintenance document content related to the user input question obtained from the stock knowledge base module and the answer to the user input question in the refined sorting module, update the answer to the user input question, and feed it back to the user.

7. According to claim 6, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that The stock knowledge base module searches the stock knowledge base for the user input question, and the specific process of obtaining the operation and maintenance document content related to the user input question is: use the NER model to perform entity query on the user input question, extract the keywords of the user input question, classify the keywords according to semantic categories, and convert the keywords into structured data, call the retrieval interface of the stock knowledge base module, and match the operation and maintenance document content related to the user input question from the stock knowledge base.

8. According to claim 1, a query-enhanced operation and maintenance knowledge base question-answering system, It is characterized in that The request format of the user input question is in XML format or JSON format, and supports jsonp cross-domain access.

9. A question-answering method for an operation and maintenance knowledge base question-answering system based on query enhancement technology according to any one of claims 1 to 8, It is characterized in that The specific steps include: Step S1: The operation and maintenance documents uploaded by the user are segmented by the data processing structured segmentation module, the segmented operation and maintenance document fragments are segmented by the feature extraction module, and the segmented operation and maintenance document fragments are vectorized using the embedding model and stored in the vector database module; Step S2: The query enhancement module expands and enhances the user input question to obtain more operation and maintenance document fragments related to the user input question; Step S3: the feature extraction module is used to perform word segmentation on all operation and maintenance document fragments related to the user input question; Step S4: The sorting module matches the segmented words with the operation and maintenance document fragments stored in the vector database module, and sorts the matched operation and maintenance document fragments in order of matching degree, and feeds back to the user as the answer to the user input question.

10. The question-answering method of the operation and maintenance knowledge base question-answering system based on query enhancement technology according to claim 9, It is characterized in that If there is an existing knowledge base module, use the NER model to perform entity query on the user input question, extract the keywords of the user input question, classify the keywords according to semantic categories, and convert the keywords into structured data, call the retrieval interface of the existing knowledge base module, match the operation and maintenance document content related to the user input question from the existing knowledge base module, and use the corpus splicing module to perform prompt splicing on the operation and maintenance document content related to the user input question obtained from the existing knowledge base and the answer to the user input question output by the refinement module, update the answer to the user input question, and feedback to the user.

Citation Information

Patent Citations

  • Information query method and device and electronic equipment

    CN111506705A

  • Search question-answering system and method based on large model and electronic equipment

    CN117708274A

  • Knowledge retrieval enhancement-based large language model question and answer method and device

    CN118113836A

  • RAG retrieval optimization method and system based on title enhancement and intention recognition

    CN118467681A