Knowledge filing method, device and equipment, storage medium and computer program product

Abstract

The invention discloses a knowledge archiving method, device and equipment, a storage medium and a computer program product, and relates to the technical field of data processing, the disclosed knowledge archiving method comprises the following steps: performing service type classification on an unstructured target demand document by adopting a vector machine algorithm to obtain a first classification result; a basic large model is adopted to generate a document abstract of the target demand document, service type classification is conducted on the target demand document according to the document abstract, a second classification result is obtained, and the basic large model is a large language model obtained through training of historical demand documents and historical document abstracts; determining a target classification result of the target demand document based on the first classification result and the second classification result; and filing the target demand document, the document abstract and the target classification result into a preset knowledge base in a structured form. According to the method and the device, the unstructured business requirement document can be efficiently archived into the structured document knowledge.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to knowledge archiving methods, devices, equipment, storage media and computer program products. Background Art

[0002] With the continuous development of corporate business and the deepening of informatization construction, business requirements documents, as an important part of corporate knowledge, record in detail the entire process and current status of corporate business establishment and development, which is crucial to improving business expansion and support capabilities.

[0003] At present, the knowledge archiving methods for business requirement documents mainly include directly archiving the requirement baseline, converting the requirement document into a structured business baseline, and automatically converting the business baseline based on a structured requirement template. However, since the information filled in by business personnel in the requirement document is usually unstructured and the form of expression is not fixed, when knowledge organizers archive and manage knowledge based on the document, the document content is difficult to convert into structured document knowledge that can be flexibly applied, which often requires a lot of manpower costs and is difficult to ensure its accuracy.

[0004] In summary, how to efficiently archive unstructured business requirement documents into structured document knowledge has become a technical problem that needs to be solved urgently in this field. Summary of the invention

[0005] The main purpose of this application is to provide a knowledge archiving method, apparatus, device, storage medium and computer program product, which aims to efficiently archive unstructured business requirement documents into structured document knowledge.

[0006] To achieve the above purpose, this application proposes a knowledge archiving method, which includes:

[0007] The vector machine algorithm is used to classify the business type of the unstructured target demand document to obtain the first classification result;

[0008] Generate a document summary of the target demand document using the basic large model, and classify the target demand document into business types according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries;

[0009] Determine a target classification result of the target requirement document based on the first classification result and the second classification result;

[0010] Archive the target requirement documents, document summaries and target classification results in a structured form into the preset knowledge base.

[0011] In one embodiment, the knowledge base includes a vector library, and the vector machine algorithm is used to classify the business type of the unstructured target demand document. The step of obtaining the first classification result includes:

[0012] After performing word segmentation and document segmentation operations on the target demand document, a vector machine algorithm is used to vectorize the target demand document after the word segmentation and document segmentation operations to obtain a structured document vector;

[0013] Performing similarity matching between the structured document vector and the historical document vector in the vector library, so as to determine the target historical demand document whose vector similarity is higher than a preset first similarity threshold from the vector library;

[0014] The business type of the target historical demand document is obtained, and the business type is used as the first classification result of the target demand document.

[0015] In one embodiment, the knowledge base includes a large model, and after the step of determining a target classification result of the target requirement document based on the first classification result and the second classification result, the step further includes:

[0016] Based on the target classification results, search the knowledge base for historical target requirement documents with the same business type as the target requirement document, and analyze the differences between the target requirement document and the historical target requirement documents to determine the business change content in the knowledge base;

[0017] Obtain business questions and standard answers corresponding to the document summary from the preset question list library;

[0018] Input the business change content into the previous version of the big model for model training to obtain the initial updated big model;

[0019] Ask business questions about the initial update model and get output answers;

[0020] Based on the comparison results between the output answers and the standard answers, the target update big model is obtained, among which the previous version big model, the basic big model and the target update big model are big models of different update rounds. The target update big model is used for automatic archiving of subsequent requirement documents and knowledge question and answer services.

[0021] In one embodiment, the step of determining a target classification result of the target requirement document based on the first classification result and the second classification result includes:

[0022] Mutually verifying the first classification result and the second classification result;

[0023] When the first classification result and the second classification result are consistent, determining that the target classification result of the target requirement document is consistent with the first classification result and the second classification result;

[0024] When the first classification result and the second classification result are inconsistent, a verification window is generated to receive a user operation, a target classification result is determined according to the user operation, and the first classification result and / or the second classification result is corrected based on the target classification result.

[0025] In one embodiment, the step of filing the target requirement document, the document summary, and the target classification result in a structured form into a preset knowledge base includes:

[0026] Build a document index of target requirement documents based on target classification results;

[0027] The target requirement documents, document indexes and target classification results after vectorization are archived into the vector library;

[0028] Archive target requirement documents, document summaries, and target classification results into the big model;

[0029] After the step of archiving the target requirement document, document summary and target classification results in a structured form into a preset knowledge base, it also includes:

[0030] Continuously compare and verify the archived knowledge of the large model and vector library.

[0031] In one embodiment, the method further comprises:

[0032] After receiving the content query instruction input from the outside, identifying the user intention corresponding to the content query instruction;

[0033] A similarity algorithm is used to select target content in the knowledge base and output the target content, wherein the target content is content whose similarity to the user's query intention is higher than a preset second similarity threshold.

[0034] In addition, to achieve the above purpose, the present application also proposes a knowledge archiving device, which includes:

[0035] A first classification module is used to classify the unstructured target demand document into business types using a vector machine algorithm to obtain a first classification result;

[0036] The second classification module is used to generate a document summary of the target demand document using the basic large model, and classify the target demand document into business types according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries;

[0037] A classification result determination module, used to determine a target classification result of the target requirement document based on the first classification result and the second classification result;

[0038] The archiving module is used to archive the target requirement documents, document summaries and target classification results in a structured form to a preset knowledge base.

[0039] In addition, to achieve the above objectives, the present application also proposes an electronic device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the knowledge archiving method as described above.

[0040] In addition, to achieve the above-mentioned purpose, the present application also proposes a storage medium, which is a computer-readable storage medium. A computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the knowledge archiving method as described above are implemented.

[0041] In addition, to achieve the above-mentioned purpose, the present application also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the steps of the knowledge archiving method as described above are implemented.

[0042] The present application proposes a knowledge archiving method. The present application first uses a vector machine algorithm to classify the business type of an unstructured target demand document to obtain a first classification result; uses a basic large model to generate a document summary of the target demand document, and based on the generated document summary, classifies the target demand document again into a business type to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries; then, the target classification result of the target demand document is determined in combination with the first classification result and the second classification result, and after determining the target classification result, the target demand document, the document summary and the target classification result are archived together in a structured form into a preset knowledge base.

[0043] In summary, this application can more accurately understand the content of unstructured target demand documents by combining vector machine algorithms and large model-based document summary generation, thereby obtaining more accurate document classification results. At the same time, the target demand documents, document summaries and classification results are stored in a structured form in the knowledge base to facilitate subsequent retrieval, query and utilization. In this way, this application achieves the effect of efficiently archiving unstructured business demand documents into structured document knowledge, which helps enterprises build a complete knowledge management system and improve the overall knowledge management level and capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0046] Figure 1 A flowchart of the first embodiment of the knowledge archiving method of the present application is provided;

[0047] Figure 2 A schematic diagram of a vector library construction process provided in Example 2 of the knowledge archiving method of this application;

[0048] Figure 3 A schematic diagram of the word segmentation and document segmentation process provided in Example 2 of the knowledge archiving method of this application;

[0049] Figure 4 A schematic diagram of an indexing process provided for Embodiment 2 of the knowledge archiving method of this application;

[0050] Figure 5 A schematic diagram of a large model optimization process provided in Example 2 of the knowledge archiving method of this application;

[0051] Figure 6 A schematic diagram of the comparison and verification process of a large model and a vector library provided in Example 2 of the knowledge archiving method of this application;

[0052] Figure 7 A schematic diagram of a large model construction framework provided in Example 2 of the knowledge archiving method of this application;

[0053] Figure 8 A schematic diagram of a vector library construction framework provided in Example 2 of the knowledge archiving method of this application;

[0054] Fig. 9 A schematic diagram of a comparison and verification framework provided for Example 2 of the knowledge archiving method of this application;

[0055] Fig.10 A schematic diagram of a search framework based on vector similarity provided in Example 2 of the knowledge archiving method of this application;

[0056] Fig.11 A schematic diagram of a search framework based on a large model question-answering robot provided in Example 2 of the knowledge archiving method of this application;

[0057] Fig.12 A schematic diagram of the knowledge archiving process provided for Embodiment 2 of the knowledge archiving method of this application;

[0058] Fig.13 This is a schematic diagram of the module structure of the knowledge archiving device according to an embodiment of the present application;

[0059] Fig.14 A schematic diagram of the device structure of the hardware operating environment involved in the knowledge archiving method in the embodiment of the present application.

[0060] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0061] It should be understood that the specific embodiments described herein are only used to explain the technical solutions of the present application and are not used to limit the present application.

[0062] In order to better understand the technical solution of the present application, a detailed description will be given below in conjunction with the accompanying drawings and specific implementation methods.

[0063] The main solution of the embodiment of the present application is: using a vector machine algorithm to classify the business type of unstructured target requirement documents to obtain a first classification result; using a basic large model to generate a document summary of the target requirement document, and classifying the target requirement document into business types based on the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical requirement documents and historical document summaries; determining a target classification result of the target requirement document based on the first classification result and the second classification result; and archiving the target requirement document, the document summary and the target classification result in a structured form into a preset knowledge base.

[0064] According to the existing technology, the knowledge archiving methods for business requirement documents mainly include directly archiving the requirement baseline, converting the requirement document into a structured business baseline, and automatically converting the business baseline based on a structured requirement template. However, since the information filled in by business personnel in the requirement document is usually unstructured and the form of expression is not fixed, when knowledge organizers archive and manage knowledge based on the document, it is difficult to convert the document content into structured document knowledge that can be flexibly applied, which often requires a lot of manpower costs and is difficult to ensure its accuracy.

[0065] The embodiment of the present application provides a solution that can more accurately understand the content of unstructured target demand documents by combining vector machine algorithms and document summary generation based on large models, thereby obtaining more accurate document classification results. At the same time, the target demand documents, document summaries and classification results are stored in a structured form in the knowledge base to facilitate subsequent retrieval, query and utilization. In this way, the present application achieves the effect of efficiently archiving unstructured business demand documents into structured document knowledge, which helps enterprises build a complete knowledge management system and improve the overall knowledge management level and capabilities.

[0066] It should be noted that the execution subject of this embodiment can be a computing service device with data processing, network communication and program running functions, such as a tablet computer, a personal computer, a mobile phone, etc., or an electronic device capable of realizing the above functions, etc. The following takes the knowledge archiving terminal as an example to illustrate this embodiment and the following embodiments.

[0067] It should be noted that, in order to facilitate understanding and explanation of this embodiment, the following first introduces the traditional knowledge archiving solution within the enterprise, as follows:

[0068] For large-scale information-based enterprises, the largest part of knowledge management and knowledge accumulation and archiving is the knowledge management and accumulation of demand documents, business plan documents, etc. during the informationization process. These documents contain the entire process and status of the establishment and development of the enterprise's business. Capturing knowledge from demand and business documents for sorting and accumulation is the most effective means of knowledge management and accumulation.

[0069] Currently, there are two main ways to accumulate and archive knowledge in requirement documents or business documents.

[0070] The first one: requirement baseline as knowledge.

[0071] A requirement baseline is a version of a requirement specification that has been confirmed and approved during the project or product development process. It represents a set of requirements that is recognized as accurate, complete, and feasible at a specific point in time. The requirement baseline serves as a foundation and reference for project or product development.

[0072] This knowledge archiving method directly archives the requirement baseline in the knowledge base, that is, directly stores the confirmed and approved requirement documents. At the same time, when archiving the requirement documents, they can be classified and stored in different directories according to business characteristics or development methods, or business tags can be added to the documents to provide general document retrieval based on file names, tags, categories, and directories. Some more advanced knowledge bases support word segmentation and word segmentation-based indexing when archiving documents, providing full-text content retrieval of the document library.

[0073] This way of knowledge archiving has the advantages of saving time and effort, but it is difficult to guarantee the validity and uniqueness of knowledge when directly archiving unstructured knowledge such as documents. It is also difficult to handle content and conflicts in different documents.

[0074] The second type: converting the requirements document into a structured business baseline.

[0075] The business baseline here refers to the business set that is abstracted by sorting out and summarizing the requirement documents of different versions and stages from the beginning of information construction to the current time.

[0076] This knowledge archiving method adds a sorting process to the knowledge accumulation process of the requirement baseline document. First, the initial version of the business baseline is established, which contains some knowledge of existing businesses. Later, in the process of requirement development, business knowledge is continuously summarized from new requirement documents. For the existing parts in the business baseline, updates and version maintenance are carried out; for business knowledge that is not in the business baseline, the first version of the new business point is established.

[0077] The advantage of this knowledge archiving method is that it can well organize the structure of the knowledge base and can accurately control the quality of the knowledge base. The disadvantage is that the workload of organizing knowledge is very large, which is not worth the effort.

[0078] The third type: Automatically convert business baselines based on structured demand templates.

[0079] This knowledge archiving method is based on the idea of ​​converting demand documents into structured business baselines. Structured templates are standardized in advance, and business personnel are asked to use templates to write and submit demand content when requirements are raised. This ensures that the demand content has a certain degree of structure, thus saving the workload of organizing business baselines from demand documents in the future.

[0080] The advantage of this method is that it greatly reduces the workload of knowledge sorting while ensuring the knowledge structure and quality of the knowledge base. The disadvantage is that the workload of sorting out templates is relatively large, and there are many restrictions when business personnel use templates to submit requirements. These technical restrictions are likely to affect business development.

[0081] In the knowledge management system of large enterprises with constantly developing businesses and frequent updates, several current methods have some shortcomings and technical problems.

[0082] The first one is the disadvantage of the demand baseline as a knowledge solution.

[0083] The method of directly archiving the requirement documents as knowledge is relatively simple. However, it has the following three disadvantages:

[0084] First, it can only be used for reading and cannot be expanded based on unstructured knowledge.

[0085] Second, the quality of knowledge is difficult to guarantee, and some knowledge is difficult to understand accurately.

[0086] Third, conflicts between different versions are difficult to identify.

[0087] Second: Disadvantages of the solution of converting requirement documents into structured business baselines.

[0088] Adding a sorting process to the knowledge archiving process, sorting unstructured documents into structured content for knowledge storage, greatly improves the accuracy of knowledge. It can also avoid the conflict handling of multiple versions and multiple changes in the first method. The main disadvantages are as follows:

[0089] First, the mapping relationship between the requirements document and business knowledge is too complicated.

[0090] Second, the quality of requirement documents affects the results of knowledge archiving.

[0091] Third, the knowledge organizers are required to have a high level of business understanding and knowledge breadth.

[0092] The third type: Disadvantages of the solution of automatically converting business baselines based on structured requirement templates.

[0093] This method of standardizing the business requirements writing process through templates is currently a mainstream development direction in the industry. This solution is equivalent to semi-structured standardization of business requirements documents. It can solve the problem of non-standard content of requirements documents to a certain extent and reduce the cost of arranging business requirements documents into structured business knowledge, but it has the following disadvantages:

[0094] First, the template has certain restrictions on business development.

[0095] Second, it is difficult to ensure through technical means that business personnel strictly follow the format.

[0096] In summary, the shortcomings of the above solutions are fundamentally due to the fact that the information in the requirements document is unstructured and its form of expression is not fixed. When conducting knowledge management, these contents are difficult to convert into structured knowledge that can be flexibly applied, which often requires a lot of manpower costs and it is difficult to ensure its accuracy.

[0097] This embodiment mainly solves the problem of how to convert unstructured documents into knowledge and improve their accuracy with lower labor costs without forcing the business requirement document content to be filled in in a structured manner.

[0098] Based on this, the present application embodiment provides a knowledge archiving method, referring to Figure 1 , Figure 1 This is a flow chart of the first embodiment of the knowledge archiving method of the present application.

[0099] In this embodiment, the knowledge archiving method includes steps S10 to S40:

[0100] Step S10, using a vector machine algorithm to classify the unstructured target demand document into business types, and obtaining a first classification result;

[0101] It should be noted that the support vector machine (SVM) algorithm is a supervised learning algorithm, which is mainly used for data classification problems. In this embodiment, through the trained vector machine, the terminal can automatically analyze the content of the unstructured target demand document, and perform preliminary classification based on the subject, field or other characteristics of the document to obtain the first classification result.

[0102] The knowledge archiving terminal uses a vector machine algorithm to perform preliminary classification of unstructured target demand documents and obtains a preliminary document classification result, namely the first classification result, wherein the target demand document is a newly created business demand document. Unstructured demand documents are difficult to be retrieved and indexed by automated tools. If the demand documents are directly archived in an unstructured form, information conflicts and inconsistencies are likely to occur.

[0103] Step S20, using the basic large model to generate a document summary of the target demand document, and classifying the target demand document into business types according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries;

[0104] It should be noted that the basic big model is a private big language model trained based on the open source big language model using historical business requirement documents and sorted historical document summaries.

[0105] The basic large model is used to generate a document summary of the target demand document, and based on the generated document summary, the target demand document is classified into business types to obtain the second classification result, wherein the document summary is a refinement and summary of the content of the target demand document, which can reflect the main content and core information of the target demand document.

[0106] Step S30, determining a target classification result of the target requirement document based on the first classification result and the second classification result;

[0107] After obtaining the first classification result and the second classification result, the two results are compared and verified. By comprehensively considering the two classification results, a more accurate and reliable target classification result is obtained to ensure the accuracy of the classification.

[0108] Step S40, archiving the target requirement document, document summary and target classification result in a structured form into a preset knowledge base.

[0109] The target requirement documents, their corresponding document summaries and the determined target classification results are archived in a structured form into the preset knowledge base. Structured archiving means that this information will be stored in the knowledge base in an orderly, easy-to-retrieve and manageable manner to facilitate subsequent use and query. At the same time, archiving documents, summaries and classification results together can provide users with comprehensive document information and convenient retrieval services.

[0110] For example, the terminal receives a newly submitted demand document, and the content of the demand document involves changes in employee welfare policies. The demand document is classified into business types through the vector machine algorithm to obtain a first classification result, and the demand document is classified as human resource management; at the same time, the document summary of the demand document is generated using the basic large model, and the demand document is classified into business types to obtain a second classification result, and the demand document is classified as employee welfare policies. Then, the first classification result and the second classification result are compared. If the two are consistent, the target classification result of the demand document can be directly determined. If the two are inconsistent, further analysis and determination are required.

[0111] In this example, since the first classification result and the second classification result are inconsistent, after further analysis it is determined that the correct classification result should be the employee welfare policy. Therefore, the target classification result of the requirement document is determined to be the employee welfare policy, and the requirement document, document summary and target classification result are archived in the knowledge base.

[0112] This embodiment provides a knowledge archiving method, which can more accurately understand the content of unstructured target demand documents by combining vector machine algorithms and document summary generation based on large models, thereby obtaining more accurate document classification results. At the same time, the target demand documents, document summaries and classification results are stored in a structured form in the knowledge base, which is convenient for subsequent retrieval, query and utilization. In this way, this application achieves the effect of efficiently archiving unstructured business demand documents into structured document knowledge, which helps enterprises build a complete knowledge management system and improve the overall knowledge management level and capabilities.

[0113] Based on the first embodiment of the present application, in the second embodiment of the present application, the same or similar contents as those in the above-mentioned embodiment 1 can be referred to the above introduction, and will not be repeated later. On this basis, the knowledge base includes a vector library, and step S10 may include steps S101 to S103:

[0114] Step S101, after performing word segmentation and document segmentation operations on the target demand document, a vector machine algorithm is used to perform vectorization processing on the target demand document after the word segmentation and document segmentation operations to obtain a structured document vector;

[0115] The knowledge archiving terminal first performs word segmentation and document segmentation operations on the target demand document to decompose the document content into smaller, meaningful units. Then, a vector machine algorithm is used to vectorize the word segmented and segmented document content to convert the target demand document from text information into a structured document vector.

[0116] Step S102, performing similarity matching between the structured document vector and the historical document vector in the vector library, so as to determine the target historical demand document whose vector similarity is higher than a preset first similarity threshold from the vector library;

[0117] The converted structured document vector is matched with the historical document vector in the vector library in the knowledge base for similarity matching to compare the similarity between the structured document vector of the target demand document and the historical document vectors in the vector library, thereby determining the target historical demand document whose vector similarity with the target demand document is higher than a first similarity threshold, wherein the first similarity threshold can be set and adjusted based on actual application requirements, and is not specifically limited in this embodiment.

[0118] Step S103, obtaining the business type of the target historical demand document, and taking the business type as the first classification result of the target demand document.

[0119] After finding the target historical requirement document with high similarity to the target requirement document through similarity matching, the business type of the target historical requirement document is obtained, and the business type is used as the first classification result of the target requirement document.

[0120] For example, to help understand this embodiment, please refer to Figure 2 , Figure 2 A simplified flow chart of building a vector knowledge database (i.e., vector library) is provided. The chart describes the process of initially building a vector library by a terminal. The vector knowledge database is built by using word segmentation and vectorization technology to vectorize knowledge content and construct an efficient index for the vectorized knowledge library to achieve high-speed retrieval of knowledge corresponding document content. The following steps are specifically included:

[0121] Sub-step 1: batch read text information of a batch of requirement documents.

[0122] Sub-step 2, word segmentation and segmentation of documents, to analyze, segment and split the unstructured knowledge information into structured forms. Among them, the word segmentation process is to segment the text content in the historical demand documents or in the large text fields of the database (such as the demand description field of the historical demand). The more mature Jieba tool (a Chinese word segmentation tool) is used, and the tool is partially customized. Parts of speech are tagged while segmenting, and words are labeled, such as adjectives, verbs, nouns, etc. After completing the word segmentation process, the result data is cleaned, the noise content is removed, and only the content related to the specific content of the business or demand is retained. The removed parts include personal names, pronouns, dates, stop words, modal particles, punctuation and stop words, such as "we", "the reason", "furthermore" and "in summary". Among them, there are two main ways of data cleaning, namely manual processing (this part is mainly based on various statically planned word libraries to extract key features) and code scripts (this part is mainly based on the program to dynamically and automatically identify and plan various word libraries based on the rule library to extract key features).

[0123] like Figure 3 As shown, the unstructured knowledge information is parsed, segmented and split through static planning and dynamic recognition respectively. Specifically, it includes cleaning invalid characters based on an artificial invalid character library (static planning) and a dynamic rule library of invalid characters (dynamic recognition), adding user dictionaries based on an artificial user dictionary (static planning) and a dynamic rule library of user dictionaries (dynamic recognition), processing synonyms and word segmentation based on an artificial synonym dictionary (static planning) and a dynamic rule library of synonyms (dynamic recognition), removing stop words based on an artificial stop word library (static planning) and a dynamic rule library of stop words (dynamic recognition), and removing non-focus words based on an artificial reserved word library (static planning) and a dynamic rule library of reserved words (dynamic recognition), and obtaining the final output feature words for subsequent document vectorization representation.

[0124] Sub-step three: Based on word segmentation and vector machine algorithms, document vectorization representation and classification are performed.

[0125] The vector similarity distance (indicating the degree of similarity) of the vectorized document content is calculated through the vector machine calculation method. Based on cluster analysis, the historical documents are classified, and the business classification is manually verified to ensure the accuracy of the basic business classification. The accurate business classification is used as the basis (it should be consistent with the business classification of the large model).

[0126] Among them, when there is no basic business classification, cluster analysis can be performed on historical demand classification, and manual classification can be performed based on the clustering results to build the initial business classification, which can save the process of manually reading each file for classification. If the project has accumulated basic business classification, the project's historical accumulated types can also be used to mark the business type of each file to obtain more accurate classification results.

[0127] Based on the results of word segmentation and cleaning, SVM is used to classify documents.

[0128] The classification algorithm uses a multi-class text classification support vector machine algorithm, which has advantages in small sample, nonlinear and high-dimensional pattern recognition. SVM is originally a binary classification algorithm. When dealing with multi-classification problems, it is necessary to construct a multi-class classifier. During training, a certain type of business description is classified as a positive set, and the remaining types of business descriptions are classified as negative sets for training. For example, if 50 categories are defined corresponding to existing business types, 50 classifiers will be obtained as the initial classification model. For subsequent newly added business requirement documents, these classification models can be used for classification. The main function of classification is to compare and verify with the business mapping results of the large model and automatically correct errors.

[0129] Sub-step four, perform distance similarity matching on the result of vectorization of the new requirement document and the existing vectors in the vector library, find the historical requirement document with high similarity (such as similarity > 70%), obtain the business type associated with the file list (i.e., historical requirement file with high similarity), and use the business type as the business type associated with the new requirement document, or as the business classification mark after the new requirement document is vectorized.

[0130] In sub-step five, the business classification tags of vectorized matching are compared with the document business classification tags of large model analysis to verify their accuracy.

[0131] Sub-step six: If the business classification tags of vectorized matching do not match the business classification tags of large model analysis, manual verification is required to mark the correct results, synchronize the tags to the vector library after marking, and serve as optimized training data for the large model.

[0132] In sub-step seven, the correctly classified data matched in sub-step five and manually marked in sub-step six are put into a vector database. The vector database can be implemented using Milvus vector library (a vector database), which can store large-scale vectorized data and has powerful retrieval and configuration functions.

[0133] Sub-step 8: Document classification and document vectorization index construction. Use Faiss (Facebook AI Similarity Search, a library for large-scale vector retrieval) indexing technology to index the document vectorization database.

[0134] Among them, Faiss is a vector-based similarity search technology. This application vectorizes the historical documents and business knowledge to construct the Faiss index.

[0135] For the newly added business requirement documents after word segmentation, Faiss is also used to first perform similarity search in the existing vector library to determine the classification that best matches the similarity with the newly added business requirement documents.

[0136] The document summary output by the large model and the corresponding business information mapping relationship are also searched for similarity through the Faiss index to find the classification with the highest similarity to the output results of the large model.

[0137] Compare the two classification results, and use this method to replace manual verification, saving labor costs.

[0138] Sub-step nine, similarity matching retrieval based on vector distance. Provide vector index matching function, and perform the same word segmentation rules on user questions and vectorize them through SVM algorithm. For example, the user wants to search for "including all changes to a certain package". During the search process, for the question input by the user, "including all changes to a certain package", first use the same method to segment the content of this sentence, remove noise words and use SVM algorithm to vectorize it, and then perform similarity matching in the vector library that has been indexed by Faiss to quickly obtain the content you want to retrieve. Implement similarity matching retrieval based on vector distance.

[0139] like Figure 4 As shown, the terminal performs document content segmentation and document vectorization on the historical document and then builds the Faiss index of the historical document. Then, when a new document is received, on the one hand, efficient Faiss indexing can be performed by query vectorization on the new document and building the Faiss of the new document, and the index can be classified and compared with the index of the historical document to obtain the content to be retrieved. On the other hand, the new document can also be segmented, SVM vectorized and the Faiss of the new document can be built, and the index can be classified and compared with the index of the historical document to obtain the content to be retrieved.

[0140] In a feasible embodiment, the knowledge base includes a large model, and step S30 may further include steps A10 to A50:

[0141] Step A10, based on the target classification result, searching the knowledge base for a historical target requirement document of the same business type as the target requirement document, and analyzing the difference between the target requirement document and the historical target requirement document to determine the business change content in the knowledge base;

[0142] After determining the target classification result of the target requirement document, search the knowledge base for historical target requirement documents with the same business type as the target requirement document, analyze the differences in content between the target requirement document and the historical target requirement documents, and identify the business change content in the indication library.

[0143] For example, if employee health insurance benefits are added to the target requirements document, but not mentioned in the historical target requirements document, this will be marked as a business change. In addition, the changes can involve business processes, data structures, policy adjustments, etc.

[0144] Step A20, obtaining business questions and standard answers corresponding to the document summary from a preset question list library;

[0145] Business questions and their standard answers corresponding to the document summary are retrieved from the preset question list library. These questions and answers represent common questions and solutions in the business field and are an important reference for evaluating the performance of large models.

[0146] Step A30, inputting the business change content into the previous version of the big model for model training to obtain the initial updated big model;

[0147] The identified business changes are input into the previous version of the big model for model training to generate an initial updated big model, thereby integrating new business knowledge and changes into the big model.

[0148] Step A40, asking business questions to the initial update large model and obtaining output answers;

[0149] After obtaining the initial updated large model, business questions are asked to the large model. The questions are generated by the question list library and are used to test the initial updated large model's ability to understand and answer business questions. The terminal can evaluate based on the output answers given by the model.

[0150] Step A50, based on the comparison result between the output answer and the standard answer, the target update big model is obtained, wherein the previous version big model, the basic big model and the target update big model are big models of different update rounds, and the target update big model is used for automatic archiving of subsequent demand documents and knowledge question and answer services.

[0151] It should be noted that the basic big model uses the internal historical demand documents of the enterprise and the historical document summaries compiled by historical business personnel based on these documents as training data. The knowledge base big model is obtained after model training. The previous version of the big model is a model version that has been further optimized and trained based on the basic big model. In actual application scenarios, the basic big model can also be updated.

[0152] The output answers of the initial update big model are compared with the standard answers to evaluate the accuracy of the initial update big model's answers to business questions, and the initial update big model is adjusted and optimized as necessary based on the comparison results to obtain the target update big model. The target update big model is based on the previous version of the big model, after merging the business change content and testing the model performance. It represents the optimal state of the big model in the current business environment and knowledge status.

[0153] In this way, based on the processing of unstructured demand documents by the big model, it can understand documents of various levels of detail, ideas, and formats, and can handle complex mapping relationships as accurately as possible to obtain more accurate archived content.

[0154] In a feasible embodiment, step S30 may include steps S301 to S303:

[0155] Step S301, mutually verifying the first classification result and the second classification result;

[0156] After obtaining the first classification result and the second classification result, the knowledge archiving terminal first verifies the two classification results against each other to ensure the consistency and accuracy between the two classification results.

[0157] Step S302, when the first classification result and the second classification result are consistent, determining that the target classification result of the target requirement document is consistent with the first classification result and the second classification result;

[0158] If the first classification result and the second classification result are consistent after mutual verification, the knowledge archiving terminal will directly determine that the target classification result of the target requirement document is the same as the two classification results.

[0159] Step S303, when the first classification result and the second classification result are inconsistent, generate a verification window to receive a user operation, determine a target classification result according to the user operation, and modify the first classification result and / or the second classification result based on the target classification result.

[0160] If the first classification result and the second classification result are inconsistent, the manual verification mechanism is triggered. By generating a verification window, the target requirement document and its two inconsistent classification results are displayed to the manual verifier for verification through the verification window. The manual verifier reclassifies the target requirement document based on his or her own professional knowledge and experience, and performs corresponding user operations, such as clicking to select the correct document type classification option in the verification window, so as to determine the target classification result.

[0161] After determining the target classification result, the knowledge archiving terminal will also correct the first classification result and / or the second classification result according to the result of manual verification to ensure the consistency and accuracy of the classification results in the knowledge base.

[0162] For example, to help understand this embodiment, please refer to Figure 5 , Figure 5 A brief process diagram for precipitating and verifying new demand business knowledge based on a big model is provided. The summary and associated business type are generated through the text data of the demand document, and the historical question library is obtained. Questions and answers are conducted on the trained big model knowledge base containing the new batch of demand documents. After automatic and manual verification, a new version of the big model is generated. Through this process, a continuously optimized big model knowledge base is constructed, which specifically includes the following steps:

[0163] Sub-step 1: batch read text information of a batch of requirement documents.

[0164] Sub-step two, generate text summaries based on the big model, extract key information based on the input document content, obtain summary information of these documents through the big model, and determine which business types in the knowledge base are changed by each requirement document (associate documents with business types).

[0165] When processing a large number of documents, the summary information of the document can help users understand the main content and key information of the document more quickly. Through the trained large model, the characteristics of the enterprise business and the different descriptions of vocabulary can be understood based on the characteristics of historical data, so as to identify the content related to the enterprise business in the newly added business requirements documents and summarize and summarize them.

[0166] The trained large model analyzes and models the document while reading the file content, determining which business types already exist in the knowledge base the document is related to and what changes have been made to the business.

[0167] Based on the big model, the document is analyzed and modeled, and the business-related content in the document is determined and mapped according to historical data and known business. In addition, the big model can also perform topic modeling and text classification to understand the subject and content of the document. By performing semantic analysis and classification on the text, the entities in the document are identified and the relationships between them are extracted, that is, structured business knowledge, or business baseline, is generated.

[0168] Sub-step three, based on the document summary information output by the large model and the business type corresponding to the document, obtain the pre-summarized business questions and standard answers related to these summary information and business types from the historical question library.

[0169] Sub-step 4: Use the files in this batch to optimize the big model and generate a new version of the big model. The new version of the big model theoretically has grasped the changes made to the business by the content of the batch of requirement documents. It has the ability to answer questions about these businesses.

[0170] Sub-step five: Use the questions in the question bank to ask questions to the new version of the large model and obtain the answers output by the new version of the large model.

[0171] Sub-step 6: Compare the semantic similarity of the answers output by the big model with the standard answers. For questions with high similarity between the answers and the standard answers in the question library, it is considered that the big model has the knowledge. For questions with low similarity between the answers and the standard answers in the question library, manual review is required to confirm whether the differences are due to changes in these needs and whether these differences are definitely caused by these changes. Specifically, when it is found that the questions are different from the previously preset answers, it is manually judged whether these differences are normal. If the business has indeed changed due to the needs of this batch, resulting in changes in the answers to the questions, and this change is found to be accurate during the verification process. This proves that the big model has mastered the changes in this batch of needs.

[0172] Sub-step 7: After manual review, confirm that the optimized and trained new version of the big model has mastered the business knowledge in the new batch of requirement documents, and then merge the new version of the big model into the main process to replace the previous version. Provide knowledge base question and answer services, as well as summary generation and business type matching functions for subsequent newly added documents.

[0173] Specifically, after manual judgment, when it is found that the answers of the large model are different from the previously preset answers, it is manually judged whether these differences are normal. If the business has indeed changed due to changes in the requirements of this batch, resulting in changes in the answers to the questions, and this change is found to be accurate during the verification process, it proves that the large model has grasped the changes in this batch of requirements.

[0174] In a feasible embodiment, step S40 may include steps S401 to S403:

[0175] Step S401, constructing a document index of a target demand document according to the target classification result;

[0176] After determining the target classification results of the target demand documents, the knowledge archiving terminal will build a corresponding document index for the structured document vector based on the classification results. The document index is used to quickly retrieve and locate document data. By building a document index, the efficiency of subsequent document retrieval and management can be significantly improved.

[0177] Step S402, archiving the target requirement document, document index and target classification result after vectorization processing into a vector library;

[0178] The target requirement document after vectorization, the corresponding document index and the target classification result are archived together in the vector library. The vector library is specially used to store vector data and document index. By archiving them to the vector library, these data can be easily managed and queried.

[0179] Step S403, archiving the target requirement document, document summary and target classification result into the big model;

[0180] In addition to archiving to the vector library, the knowledge archiving terminal will also archive the target requirement document, its document summary and the target classification results into the big model. The big model is a trained and optimized large language model that can process and analyze large amounts of data and generate corresponding results.

[0181] By archiving the target requirement document and its related information into the big model, the big model can be fully utilized to conduct in-depth analysis and processing of the document.

[0182] In a feasible embodiment, step S40 may further include step B10:

[0183] Step B10, continuously comparing and verifying the archived knowledge of the large model and the vector library.

[0184] In order to ensure the accuracy and consistency of archived knowledge, the knowledge archiving terminal will continuously compare and verify the knowledge in the large model and vector library to ensure the consistency and accuracy of the knowledge between the two. If inconsistencies or errors are found in the knowledge, they will be corrected and updated in a timely manner to ensure the reliability and accuracy of the archived knowledge.

[0185] For example, to help understand this embodiment, please refer to Figure 6 , Figure 6 A process for comparing and verifying a large model and a vector library is provided, which specifically includes the following steps:

[0186] Sub-step 1: Construction of large model knowledge base.

[0187] like Figure 7As shown, in the process of building the knowledge base big model, open source big models such as GPT3.0 (General Pre-trained Transformer-3 third-generation general pre-trained converter) or ChatGLM (Chinese-language Pre-training Model, an open source big language model) can be used as the basic big model, and the internal historical business requirement documents of the enterprise and the historical document summaries and business list information sorted out by historical business personnel based on these documents are used as training data. The knowledge base big model obtained after model training has the ability to obtain summaries from document texts in specified business fields.

[0188] Sub-step 2: Building document vectorization indexing and classification capabilities.

[0189] like Figure 8 As shown, the core of the construction process of document vectorization index and classification capability is the tuning of word segmentation granularity, splitting granularity and classification algorithm. In this embodiment, historical business demand documents are used as input, and historical document summaries and business lists sorted out by business personnel on the basis of these documents and document and business classification mapping are used as basic information for output verification, and the parameters and granularity of the corresponding stages are tuned, that is, the word segmentation and document segmentation, document vectorization index and SVM vector machine classification algorithm in the vectorization index and classification tool are tuned, until the output file of the whole process is consistent with the content sorted by the business personnel, then the construction of document vectorization index and classification capability is successful.

[0190] Sub-step three: Mutual verification of the large model inference classification results and the vectorized index classification results.

[0191] like Fig. 9 As shown, in this embodiment, when a new business requirement document is input, the reasoning of the big model and the classification analysis of the similarity of the document after vectorization are performed simultaneously to generate a mapping of two groups of documents and business classifications, that is, the document business change summary and business classification are obtained through the knowledge base big model, and the document and business classification similarity based on vector analysis are obtained through vectorized indexing and classification tools. Then, the two classification results are automatically compared and audited. For those that are consistent, they are directly entered into the knowledge base. For documents with large comparison differences, they are used as input for manual comparison and manual mapping is performed, and then the mapping results are used to optimize and train the knowledge base big model.

[0192] This continuous workflow can ensure the accuracy of the knowledge base on the one hand, and continuously optimize the large model on the other hand, and the data collection process for optimized training becomes a sustainable and low-cost task.

[0193] In a feasible embodiment, the knowledge archiving method may further include steps B50 to B60:

[0194] Step B50, after receiving the content query instruction input from the outside, identifying the user intention corresponding to the content query instruction;

[0195] When the knowledge archiving terminal receives a content query instruction input from the outside, it identifies the user intention represented by the instruction, wherein the user intention refers to the specific content of the information or answer that the user hopes to obtain from the knowledge base.

[0196] Step B60, using a similarity algorithm to select target content in the knowledge base, and output the target content, wherein the target content is content whose similarity to the user's query intention is higher than a preset second similarity threshold.

[0197] After identifying the user's intention, the knowledge archiving terminal will use the similarity algorithm to search and match in the knowledge base, and output the found target content to the user for viewing to complete the user's query request. The target content is the content whose similarity with the user's query intention is higher than the second similarity threshold. The second similarity threshold can be set and adjusted based on actual application needs, and this is not specifically limited in this embodiment.

[0198] In addition, it should be noted that there are three ways for the terminal to provide query content for externally input content query instructions. The first is a knowledge base retrieval function similar to other knowledge bases, which will not be elaborated here. The second is knowledge and business document search based on vector similarity. The third is to build a question-and-answer robot based on a large model.

[0199] Among them, knowledge and business document search functions based on vector similarity, such as Fig.10 As shown, since the documents and business knowledge are indexed by Faiss vectorization based on similarity in this embodiment, users can use similarity-based search when searching for knowledge. The query content can be a keyword or an interactive language that describes a query purpose. The query content is constructed as a Faiss index. During the search process, the basic large model is used to perform natural language semantic analysis and user intention inference on the input query. For example: input an interactive question such as "Can employees work part-time outside?" The search process does not just use "Can employees work part-time outside?" as a keyword for retrieval or full-text search. Instead, it understands that the user wants to ask about the part-time policy. After understanding the user's intention, the business needs and business knowledge content in the Faiss index library are found according to the similarity search algorithm, and the similarity of each content with the user's intention is calculated. The program can take the x contents with the best similarity for output.

[0200] Among them, the question-answering robot function based on the large model is constructed, such as Fig.11As shown, in this embodiment, a knowledge base big model that masters all the business knowledge of the knowledge base is constructed, and an interface can be provided to implement question-answering services based on the big model and knowledge base knowledge, and record usage records and user feedback information. Specifically, a robot interface embedded in the page is provided on the client, and the robot service and model are classified, and the questions and reply data generated during the query process are stored in mysql (a data management system), and based on LangChain (Language Chain is used to build applications based on big language models) and big language models, http (Hypertext Transfer Protocol) requests are completed to realize personalized knowledge question-answering services.

[0201] Thus, in this embodiment, knowledge archiving does not require the process of filling in business requirements in a standardized template. Instead, it introduces big models and machine learning technology to assist in the process of extracting business knowledge from unstructured business requirement documents, using machines instead of manual labor to implement the knowledge organization process with the highest labor cost, and manual labor only needs to verify the results.

[0202] In this embodiment, the final precipitated business knowledge is divided into three parts: a business baseline (ie, a structured business knowledge base), a document vectorization index, and a large model trained using demand documents.

[0203] The business baseline is the highly abstracted business knowledge from the requirements document, which is the overall view of the enterprise business and is structured and saved according to the business classification. It also provides corresponding search and structured display applications.

[0204] The document vectorization index is a document index formed by natural language analysis and clustering methods for all demand documents. It is a set of document classifications calculated based on natural language and vector machine algorithms. It saves the vectorization distance between documents and reflects the similarity between documents.

[0205] The big model trained with requirement documents uses historical requirement documents to train the big model technology, so that the big model can grasp the construction intention of the requirement documents, thereby understanding the business meaning of the requirements. The trained big model grasps all the knowledge in the knowledge base and can organize the knowledge content in human language. This form of knowledge enables the big model to provide knowledge query and knowledge question and answer through dialogue, question and answer, etc.

[0206] The sources of the three types of knowledge are all historical requirements documents, which are generated using different technical means. They can be continuously generated through machine learning and large model training. In addition, the three types of knowledge can verify each other's accuracy, reducing the complexity of manual evaluation of training results.

[0207] For example, to help understand the implementation process of the knowledge archiving method obtained by combining this embodiment with the above embodiments, please refer to Fig.12 , Fig.12 A brief flowchart of a knowledge archiving method is provided, specifically:

[0208] First, the newly approved new business requirement document is first segmented and documented, then vectorized and classified using a vector machine algorithm (i.e., SVM algorithm), and a document vectorization index is constructed. At the same time, a basic large model (i.e., a basic large language model) is used to generate a summary of the document business change for the requirement document. The summary of the new business requirement document is compared with the existing business classification in the library through keyword frequency, and the classification result of which type of business the new document change belongs to is output. This type of result is used as the classification tagging result of the large model for the new document, that is, the business and document classification mapping relationship is obtained. At the same time, the classification tagging result of the large model is compared with the tagging result output by the vector machine algorithm. For the parts with the same tagging results of the two methods, the new documents can be directly archived to the vector library, including the document classification and vector index library. The inconsistent parts require manual inspection and tagging.

[0209] With the continuous optimization of large models and the continuous accumulation of vector library data (including manual inspection and labeling), the classification result labeling will become more and more accurate, which will gradually reduce the workload of manual inspection and labeling of this part of the work.

[0210] Among them, the basic big model for generating document summaries and structurally transforming knowledge information in documents is a private big model trained based on the open source big model using historical business requirement documents and organized summary information and structured knowledge information. The big model is continuously optimized and trained by using the summary results and structured knowledge results that have been determined accurately in the process. The accurate information includes: the consistent parts of the output results of the big model and the vectorized output in the process, and the parts that are manually checked and marked after the outputs of the two methods are inconsistent.

[0211] At the same time, the big model has the ability to answer questions based on the knowledge base content by asking questions. In the pre-maintained business and test question list library, the question library related to the new business is queried, and the summary information of the new document generated by the big model is automatically answered based on the question library. The accuracy of the question and answer records can be manually checked to verify whether the merged big model meets expectations. The data that meets expectations is also used as the data for the optimization training of the big model. The big model's ability to generate answers is continuously optimized and trained.

[0212] There are different versions of this private big model. After optimization and training, the big model is merged with the previous version of the big model in terms of capabilities. Through continuous optimization and training, the accuracy of the big model will continue to improve with version iteration and capability merging, giving the process certain self-optimization characteristics.

[0213] It should be noted that the above examples are only used to understand the present application and do not constitute a limitation on the knowledge archiving method of the present application. More simple transformations based on this technical concept are all within the scope of protection of the present application.

[0214] The present application also provides a knowledge archiving device, please refer to Fig.13 , the knowledge archiving device includes:

[0215] A first classification module 10 is used to classify the unstructured target demand document into business types using a vector machine algorithm to obtain a first classification result;

[0216] The second classification module 20 is used to generate a document summary of the target demand document using the basic large model, and classify the target demand document into business types according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries;

[0217] A classification result determination module 30, configured to determine a target classification result of the target requirement document based on the first classification result and the second classification result;

[0218] The archiving module 40 is used to archive the target requirement document, document summary and target classification result in a structured form to a preset knowledge base.

[0219] Optionally, the knowledge base includes a vector library, and the first classification module 10 is further used for:

[0220] After performing word segmentation and document segmentation operations on the target demand document, a vector machine algorithm is used to vectorize the target demand document after the word segmentation and document segmentation operations to obtain a structured document vector;

[0221] Performing similarity matching between the structured document vector and the historical document vector in the vector library, so as to determine the target historical demand document whose vector similarity is higher than a preset first similarity threshold from the vector library;

[0222] The business type of the target historical demand document is obtained, and the business type is used as the first classification result of the target demand document.

[0223] Optionally, the knowledge base includes a large model, and the knowledge archiving device further includes a large model updating module (not shown), and the large model updating module is used to:

[0224] Based on the target classification results, search the knowledge base for historical target requirement documents with the same business type as the target requirement document, and analyze the differences between the target requirement document and the historical target requirement documents to determine the business change content in the knowledge base;

[0225] Obtain business questions and standard answers corresponding to the document summary from the preset question list library;

[0226] Input the business change content into the previous version of the big model for model training to obtain the initial updated big model;

[0227] Ask business questions about the initial update model and get output answers;

[0228] Based on the comparison results between the output answers and the standard answers, the target update big model is obtained, among which the previous version big model, the basic big model and the target update big model are big models of different update rounds. The target update big model is used for automatic archiving of subsequent requirement documents and knowledge question and answer services.

[0229] Optionally, the classification result determination module 30 is further configured to:

[0230] Mutually verifying the first classification result and the second classification result;

[0231] When the first classification result and the second classification result are consistent, determining that the target classification result of the target requirement document is consistent with the first classification result and the second classification result;

[0232] When the first classification result and the second classification result are inconsistent, a verification window is generated to receive a user operation, a target classification result is determined according to the user operation, and the first classification result and / or the second classification result is corrected based on the target classification result.

[0233] Optionally, the knowledge archiving device further includes a comparison and verification module (not shown), which is used to:

[0234] Build a document index of target requirement documents based on target classification results;

[0235] The target requirement documents, document indexes and target classification results after vectorization are archived into the vector library;

[0236] Archive target requirement documents, document summaries, and target classification results into the big model;

[0237] Continuously compare and verify the archived knowledge of the large model and vector library.

[0238] Optionally, the knowledge archiving device further includes a content query module (not shown), which is used to:

[0239] After receiving the content query instruction input from the outside, identifying the user intention corresponding to the content query instruction;

[0240] A similarity algorithm is used to select target content in the knowledge base and output the target content, wherein the target content is content whose similarity to the user's query intention is higher than a preset second similarity threshold.

[0241] The knowledge archiving device provided in the embodiment of the present application adopts the knowledge archiving method in the above embodiment, and can efficiently archive unstructured business demand documents into structured document knowledge. Compared with the prior art, the beneficial effects of the knowledge archiving device provided in the embodiment of the present application are the same as the beneficial effects of the knowledge archiving method provided in the above embodiment, and other technical features in the knowledge archiving device are the same as the features disclosed in the above embodiment method, which will not be repeated here.

[0242] An embodiment of the present application provides an electronic device, which includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the knowledge archiving method in the above-mentioned embodiment one.

[0243] Reference below Fig.14 , which shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present application. The electronic devices in the embodiments of the present application may include but are not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig.14 The knowledge archiving device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0244] like Fig.14As shown, the electronic device may include a processing device 1001 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read Only Memory) 1002 or a program loaded from a storage device 1003 to a random access memory (RAM: Random Access Memory) 1004. In RAM1004, various programs and data required for the operation of the electronic device are also stored. The processing device 1001, ROM1002, and RAM1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems can be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 can allow the electronic device to communicate with other devices wirelessly or by wire to exchange data. Although the figure shows an electronic device with various systems, it should be understood that it is not required to implement or have all the systems shown. More or fewer systems can be implemented or have alternatively.

[0245] In particular, according to the embodiments disclosed in the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the embodiments disclosed in the present application include a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a 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 1003, or installed from a ROM 1002. When the computer program is executed by the processing device 1001, the above-mentioned functions defined in the method of the embodiment disclosed in the present application are executed.

[0246] The electronic device provided in the embodiment of the present application adopts the knowledge archiving method in the above embodiment, and can efficiently archive unstructured business demand documents into structured document knowledge. Compared with the prior art, the beneficial effects of the electronic device provided in the embodiment of the present application are the same as the beneficial effects of the knowledge archiving method provided in the above embodiment, and the other technical features in the electronic device are the same as the features disclosed in the method of the previous embodiment, which will not be repeated here.

[0247] It should be understood that the various parts disclosed in the embodiments of the present application can be implemented by hardware, software, firmware or a combination thereof. In the description of the above embodiments, specific features, structures, materials or characteristics can be combined in any one or more embodiments or examples in a suitable manner.

[0248] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0249] An embodiment of the present application provides a computer-readable storage medium having computer-readable program instructions (ie, computer programs) stored thereon, and the computer-readable program instructions are used to execute the knowledge archiving method in the above-mentioned embodiment.

[0250] The computer-readable storage medium provided in the embodiment of the present application may be, for example, a USB flash drive, but is not limited to electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, systems or devices, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM: Random Access Memory), a read-only memory (ROM: Read Only Memory), an erasable programmable read-only memory (EPROM: Erasable Programmable Read Only Memory or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM: CD-Read Only Memory), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, system or device. The program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (Radio Frequency: Radio Frequency), etc., or any suitable combination of the above.

[0251] The computer-readable storage medium may be included in the electronic device, or may exist independently without being installed in the electronic device.

[0252] The above-mentioned computer-readable storage medium carries one or more programs. When the above-mentioned one or more programs are executed by an electronic device, the electronic device: uses a vector machine algorithm to classify the business type of an unstructured target demand document to obtain a first classification result; uses a basic large model to generate a document summary of the target demand document, and classifies the business type of the target demand document according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries; determines a target classification result of the target demand document based on the first classification result and the second classification result; and archives the target demand document, the document summary and the target classification result in a structured form to a preset knowledge base.

[0253] The computer program code for performing the operation of the embodiment of the present application can be written in one or more programming languages ​​or a combination thereof, and the above-mentioned programming languages ​​include object-oriented programming languages, such as Java, Smalltalk, C++, and also include conventional procedural programming languages, such as "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on the remote computer, or completely on the remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0254] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0255] The modules involved in the embodiments described in this application may be implemented by software or hardware, wherein the name of the module does not constitute a limitation on the unit itself in some cases.

[0256] The readable storage medium provided in the embodiment of the present application is a computer-readable storage medium, which stores computer-readable program instructions (i.e., computer programs) for executing the above-mentioned knowledge archiving method, and can efficiently archive unstructured business demand documents into structured document knowledge. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in the embodiment of the present application are the same as the beneficial effects of the knowledge archiving method provided in the above-mentioned embodiment, and will not be repeated here.

[0257] An embodiment of the present application also provides a computer program product, including a computer program, which implements the steps of the knowledge archiving method as described above when the computer program is executed by a processor.

[0258] The computer program product provided in the embodiment of the present application can efficiently archive unstructured business requirement documents into structured document knowledge. Compared with the prior art, the beneficial effects of the computer program product provided in the embodiment of the present application are the same as the beneficial effects of the knowledge archiving method provided in the above embodiment, which will not be repeated here.

[0259] The above are only some embodiments of the present application, and are not intended to limit the patent scope of the present application. All equivalent structural changes made using the contents of the present application specification and drawings under the technical concept of the present application, or direct / indirect application in other related technical fields are included in the patent protection scope of the present application.

Claims

1. A knowledge archiving method, characterized in that: The method includes: The vector machine algorithm is used to classify the business type of the unstructured target demand document to obtain the first classification result; Generate a document summary of the target demand document using a basic large model, and classify the target demand document into business types according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries; Determine a target classification result of the target requirement document based on the first classification result and the second classification result; The target requirement document, the document summary and the target classification result are filed in a structured form into a preset knowledge base.

2. The method according to claim 1, characterized in that The knowledge base includes a vector library, and the step of using a vector machine algorithm to classify the unstructured target demand document into business types to obtain a first classification result includes: After performing word segmentation and document segmentation operations on the target demand document, a vector machine algorithm is used to vectorize the target demand document after the word segmentation and document segmentation operations to obtain a structured document vector; Performing similarity matching between the structured document vector and the historical document vector in the vector library, so as to determine a target historical demand document having a vector similarity higher than a preset first similarity threshold from the vector library; The business type of the target historical demand document is obtained, and the business type is used as a first classification result of the target demand document.

3. The method according to claim 1, characterized in that The knowledge base includes a large model, and after the step of determining the target classification result of the target requirement document based on the first classification result and the second classification result, the step further includes: Based on the target classification result, searching the knowledge base for a historical target requirement document of the same business type as the target requirement document, and analyzing the difference between the target requirement document and the historical target requirement document to determine the business change content in the knowledge base; Obtaining business questions and standard answers corresponding to the document summary from a preset question list library; Input the business change content into the previous version of the big model for model training to obtain the initial updated big model; Asking the business question to the initial update big model and obtaining an output answer; According to the comparison result between the output answer and the standard answer, the target update big model is obtained, wherein the previous version big model, the basic big model and the target update big model are the big models of different update rounds, and the target update big model is used for automatic archiving of subsequent demand documents and knowledge question and answer services.

4. The method according to claim 1, characterized in that The step of determining the target classification result of the target requirement document based on the first classification result and the second classification result comprises: Mutually verifying the first classification result and the second classification result; When the first classification result and the second classification result are consistent, determining that the target classification result of the target requirement document is consistent with the first classification result and the second classification result; When the first classification result and the second classification result are inconsistent, a verification window is generated to receive a user operation, the target classification result is determined according to the user operation, and the first classification result and / or the second classification result is corrected based on the target classification result.

5. The method according to claim 2, characterized in that The step of filing the target requirement document, the document summary and the target classification result in a structured form into a preset knowledge base includes: Constructing a document index of the target requirement document according to the target classification result; Archiving the target requirement document, the document index and the target classification result after vectorization processing into a vector library; Archiving the target requirement document, the document summary and the target classification result into a large model; After the step of filing the target requirement document, the document summary and the target classification result in a structured form into a preset knowledge base, the method further includes: The archived knowledge of the large model and the vector library are continuously compared and verified.

6. The method according to claim 5, characterized in that The method further comprises: After receiving a content query instruction input from outside, identifying a user intention corresponding to the content query instruction; A similarity algorithm is used to select target content in the knowledge base, and the target content is output, wherein the target content is content whose similarity with the user's query intention is higher than a preset second similarity threshold.

7. A knowledge archiving device, characterized in that: The device comprises: A first classification module is used to classify the unstructured target demand document into business types using a vector machine algorithm to obtain a first classification result; A second classification module is used to generate a document summary of the target demand document using a basic large model, and classify the target demand document into business types according to the document summary to obtain a second classification result, wherein the basic large model is a large language model trained using historical demand documents and historical document summaries; A classification result determination module, used to determine a target classification result of the target requirement document based on the first classification result and the second classification result; The archiving module is used to archive the target requirement document, the document summary and the target classification result in a structured form to a preset knowledge base.

8. An electronic device, characterized in that: The device comprises: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the computer program is configured to implement the steps of the knowledge archiving method according to any one of claims 1 to 6.

9. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, the steps of the knowledge archiving method according to any one of claims 1 to 6 are implemented.

10. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the knowledge archiving method according to any one of claims 1 to 6 are implemented.

Images