Official document information security management method and system based on large language model

Abstract

The invention discloses an official document information security management method and system based on a large language model. The method comprises the steps of determining a default information base of a target user; if the matched reply data exists in the default information base, feeding back the reply data in the default information base to the target user; if the default information base does not have the matched reply data, judging whether the knowledge base has associated data or not; if yes, inputting the associated data and the user question into a large language model, outputting reply data from the large language model, and feeding back the reply data to the target user; if not, inputting the user question into the large language model, and feeding back reply data output by the large language model to the target user; the default information base and / or the knowledge base adopts a preset management algorithm for safety management, and the default information base and / or the knowledge base comprises at least one piece of official document information. According to the scheme, on the basis of guaranteeing the accuracy of the reply data of the user question, the solving efficiency of the user question is improved, and system resources are saved; and the user data security is ensured.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, specifically to a document information security management method based on a large language model, a document information security management system based on a large language model, a computing device, a computer storage medium, and a computer program product. Background Technology

[0002] A large language model is a language processing model based on artificial intelligence technology. It can provide corresponding answers to user questions, greatly facilitating people's work and life.

[0003] However, the inventors discovered the following defects in the existing technology during implementation: the existing technology only uses a large language model to process user questions without combining the user's private knowledge and data, resulting in low accuracy and insufficient processing efficiency when processing user questions; moreover, the data relied upon by the large language model in the existing technology is public data, which has poor data security. Summary of the Invention

[0004] In view of the above problems, this application is made in order to provide a document information security management method based on a large language model, a document information security management system based on a large language model, a computing device, a computer storage medium, and a computer program product that overcomes or at least partially solves the above problems.

[0005] According to the first aspect of this application, a method for managing the security of official document information based on a large language model is provided, including:

[0006] Obtain user questions provided by the target users and determine the default information database corresponding to the target users;

[0007] If there is a matching answer in the default information database, the answer in the default information database will be sent back to the target user.

[0008] If no answer data matching the user's question exists in the default information base, determine whether there is related data for the user's question in the knowledge base;

[0009] If so, input the associated data and user questions into the large language model, and feed back the response data output by the large language model to the target user;

[0010] If not, input the user's question into the large language model and feed back the response data output by the large language model to the target user;

[0011] The default information base and / or the knowledge base are managed using a preset management algorithm for security purposes, and the default information base and / or the knowledge base contain at least one official document.

[0012] In one alternative implementation, the default information base is constructed in the following manner:

[0013] Obtain the data uploaded by the target user through the default information database configuration entry, and use the uploaded data as the default information database data;

[0014] And / or, obtain the target user's historical browsing data and historical browsing frequency, and determine the default information database data from the historical browsing data based on the historical browsing frequency;

[0015] And / or, obtain the historical user question association data of the target user, and determine the default information database data from the association data based on the frequency of occurrence of the association data and the target user's response feedback;

[0016] The default information database data is stored in the default information database corresponding to the target user.

[0017] In an optional implementation, the method further includes:

[0018] Based on the hit frequency and data storage duration of the default information database, the data to be eliminated is determined from the default information database.

[0019] Remove obsolete data from the default information base and store the obsolete data in the knowledge base.

[0020] In an optional implementation, the method further includes:

[0021] Retrieve knowledge base documents;

[0022] The knowledge base documents are segmented according to their document types using a matching segmentation method to obtain at least one knowledge base data.

[0023] Store knowledge base data in the knowledge base.

[0024] In one optional implementation, determining whether the knowledge base contains associated data related to user questions includes:

[0025] If the similarity between knowledge base data and user question is greater than the first similarity threshold, then the knowledge base data will be used as the associated data of the user question.

[0026] If the similarity between knowledge base data and user question is less than or equal to the first similarity threshold and greater than the second similarity threshold, then the knowledge base data is used as candidate related data, and the related data of user question is determined from the candidate related data and the context data of the candidate related data.

[0027] If the similarity between each knowledge base entry and the user's question is less than or equal to the second similarity threshold, then it is determined that there is no related data for the user's question in the knowledge base.

[0028] The second similarity threshold is less than the first similarity threshold.

[0029] In one optional implementation, determining the associated data of the user question from the candidate associated data and the context data of the candidate associated data includes:

[0030] For any candidate related data, the candidate related data and its context data are further segmented, and the related data of the user's question are determined based on the results of the secondary segmentation.

[0031] According to a second aspect of this application, a document information security management system based on a large language model is provided, comprising:

[0032] The acquisition module is used to acquire user questions provided by the target user and determine the default information database corresponding to the target user.

[0033] The default information database module is used to send the answer data from the default information database to the target user if there is answer data in the default information database that matches the user's question.

[0034] The knowledge base module is used to determine whether there is related data for the user's question in the knowledge base if no answer data matching the user's question exists in the default information base.

[0035] The large language model module is used to input the relevant data and user questions into the large language model if the condition is met, and then return the response data output by the large language model to the target user if the condition is not met.

[0036] The default information base and / or the knowledge base are managed using a preset management algorithm for security purposes, and the default information base and / or the knowledge base contain at least one official document.

[0037] In one optional implementation, the default information base module is used to: obtain data uploaded by the target user through the default information base configuration entry, and use the uploaded data as the default information base data;

[0038] And / or, obtain the target user's historical browsing data and historical browsing frequency, and determine the default information database data from the historical browsing data based on the historical browsing frequency;

[0039] And / or, obtain the historical user question association data of the target user, and determine the default information database data from the association data based on the frequency of occurrence of the association data and the target user's response feedback;

[0040] The default information database data is stored in the default information database corresponding to the target user.

[0041] In one optional implementation, the default information base module is used to: determine the discarded data from the default information base based on the hit frequency of the default information base data and the data storage duration;

[0042] The knowledge base module is used to remove obsolete data from the default information database and store obsolete data in the knowledge base.

[0043] In one alternative implementation, the knowledge base module is used to: retrieve knowledge base documents;

[0044] The knowledge base documents are segmented according to their document types using a matching segmentation method to obtain at least one knowledge base data.

[0045] Store knowledge base data in the knowledge base.

[0046] In one optional implementation, the knowledge base module is used to: if the similarity between knowledge base data and user question is greater than a first similarity threshold, then use the knowledge base data as associated data of the user question;

[0047] If the similarity between knowledge base data and user question is less than or equal to the first similarity threshold and greater than the second similarity threshold, then the knowledge base data is used as candidate related data, and the related data of user question is determined from the candidate related data and the context data of the candidate related data.

[0048] If the similarity between each knowledge base entry and the user's question is less than or equal to the second similarity threshold, then it is determined that there is no related data for the user's question in the knowledge base.

[0049] The second similarity threshold is less than the first similarity threshold.

[0050] In one optional implementation, the knowledge base module is used to: for any candidate related data, perform a secondary segmentation on the candidate related data and its context data, and determine the related data of the user's question based on the secondary segmentation result.

[0051] According to a third aspect of this application, a computing device is provided, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other through the communication bus;

[0052] The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described document information security management method based on a large language model.

[0053] According to a fourth aspect of this application, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, the executable instruction causing a processor to perform the operation corresponding to the above-described document information security management method based on a large language model.

[0054] According to a fifth aspect of this application, a computer program product is provided, comprising at least one executable instruction that causes a processor to perform the operations corresponding to the above-described document information security management method based on a large language model.

[0055] According to the document information security management method, system, computing device, computer storage medium and computer program product based on the large language model provided in this application, the answer data for user questions is first searched from the default information database customized by the target user, and the answer data is directly returned to the user if it can be found; if the default information database cannot solve the user's question, the answer data is obtained by using the knowledge base and the large language model, thereby improving the efficiency of solving user questions and saving system resources while ensuring the accuracy of the answer data.

[0056] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0057] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0058] Figure 1 A flowchart illustrating a document information security management method based on a large language model provided in an embodiment of this application is shown.

[0059] Figure 2 A flowchart of a default information base construction method provided in an embodiment of this application is shown;

[0060] Figure 3 A flowchart of a method for finding related data provided in an embodiment of this application is shown;

[0061] Figure 4 This paper illustrates a structural diagram of a document information security management system based on a large language model, as provided in an embodiment of this application.

[0062] Figure 5A structural diagram of a computing device provided in an embodiment of this application is shown. Detailed Implementation

[0063] Exemplary embodiments of the present application will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it should be understood that the present application may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the scope of the present application to those skilled in the art.

[0064] Figure 1 A flowchart illustrating a document information security management method based on a large language model, provided in an embodiment of this application, is shown. Figure 1 As shown, the method includes the following steps:

[0065] Step S110: Obtain user questions provided by the target user and determine the default information database corresponding to the target user.

[0066] Target users can input relevant question data through text, images, voice, etc., and the corresponding user questions can be extracted by processing the question data. This invention does not limit the specific method of user question extraction; for example, semantic parsing can be performed on the user-input text to extract user questions, or non-text data such as user-input images or voice can be converted into text data, and then user questions can be extracted from that text data.

[0067] This invention provides a pre-built default information database associated with each user. This database contains data related to the user, including at least one official document, such as the company bylaws of the user's company. The invention then extracts the target user's identifier and, based on the pre-established mapping between the default information database and the user identifier, quickly determines the corresponding default information database for that user. This implementation allows for the customization of default information databases for different users, enabling privatization, exclusivity, customization, and personalization of the database, ensuring that the constructed database matches the user's individual characteristics.

[0068] Step S120: Determine whether there is matching response data in the default information database; if yes, proceed to step S130; if no, proceed to step S140.

[0069] The process compares user questions provided by the target user with data in the target user's default information base, prioritizing the search for matching response data in the default information base. In an optional implementation, to improve user question processing efficiency, both the user questions and the data in the default information base can be vectorized. By comparing the similarity between the user question vector and the data vectors in the default information base, matching response data from the default information base can be quickly and accurately determined.

[0070] If a matching response can be found in the default information database, proceed to step S130; if a matching response cannot be found in the default information database, proceed to step S140.

[0071] Step S130: Feedback the response data from the default information database to the target user.

[0072] If a matching answer exists in the default information base, the answer from the default information base will be fed back to the target user, thus eliminating the need for further processing of the knowledge base and large language model, improving the efficiency of user question processing, and saving system resources.

[0073] Step S140: Determine whether there is related data in the knowledge base; if yes, proceed to step S150; if no, proceed to step S160.

[0074] If no matching answer data is found in the default information base, further processing of the knowledge base will be performed.

[0075] Specifically, this embodiment of the invention also constructs a knowledge base containing corresponding knowledge base data, which includes at least one official document. This knowledge base data can originate from user-uploaded data, domain-specific data, etc. The data in the knowledge base is associated with the current user and can also be referred to as the user's private data, customized data, etc. This knowledge base data is typically stored in the knowledge base after preprocessing (such as format conversion, invalid field removal, etc.). Compared to the knowledge base, the default information repository contains data with higher user relevance, faster retrieval speed, and consumes fewer system resources.

[0076] In this embodiment of the invention, the default information repository and / or knowledge base can be a private repository associated with the user. Users can upload internal data, official documents, professional knowledge, and other materials from their organizations, such as their company, to build their own private and customized default information repository and / or knowledge base. A preset security management algorithm is used to manage this repository and ensure information security. This preset security management algorithm includes an access control algorithm, which can assign customized permissions to different user IDs. These permissions include: usage permissions, permissions to view internal data, permissions to modify internal data, and / or login permissions, etc. Furthermore, this default information repository and / or knowledge base can be deployed locally, on a preset website, or on a system, etc., according to user needs.

[0077] Furthermore, to further ensure data security and to manage the default information repository and / or knowledge base securely, a security protection module can be encapsulated inside or outside the default information repository and / or knowledge base. This security protection module can detect and block abnormal operation behaviors on the default information repository and / or knowledge base, including but not limited to: access by high-risk users, abnormal frequency of user access, etc.

[0078] Furthermore, to facilitate the management of the knowledge base, this embodiment of the invention also provides users with a knowledge base management portal, allowing them to categorize the data within the knowledge base. For example, knowledge base data can be divided into different categories based on data management permissions.

[0079] The vectorized user question is matched with the vectorized knowledge base data. Based on the matching result, it is determined whether there is any answer data in the knowledge base that matches the user question. If the relevant data of the user question can be found in the knowledge base, then step S150 is executed; if the relevant data of the user question cannot be found in the knowledge base, then step S160 is executed.

[0080] In one optional implementation, the system provides a knowledge base selection entry point for the user, allowing the user to select the appropriate knowledge base for processing according to their needs. If the user does not select any knowledge base, step S160 is executed directly.

[0081] Step S150: Input the associated data and user questions into the large language model, and feed back the response data output by the large language model to the target user.

[0082] If the knowledge base contains related data about user questions, then that related data and the user questions are input into the large language model. For example, related data, prompts, user questions, and historical conversation information can all be input into the large language model. The large language model can then analyze this data and output corresponding response data, thereby obtaining relevant response data using a retrieval-enhanced large language model.

[0083] In this process, historical conversations can be filtered to remove irrelevant information and extract only the valid ones (conversations that meet the user's needs), thereby improving the accuracy of the generated response data. Finally, the response data output by the large language model is fed back to the target user.

[0084] Step S160: Input the user's question into the large language model, and feed back the response data output by the large language model to the target user.

[0085] If the default information base lacks response data for the user's question, and the knowledge base lacks related data for the user's question, then the response data is obtained directly using the large language model. If the large language model determines that the current information is insufficient, the system categorizes the questions that can be resolved with existing information and the information that should be provided, and then proceeds to the next round of interaction. This embodiment of the invention does not limit the processing method of the large language model.

[0086] Therefore, this embodiment of the invention prioritizes searching for answer data for user questions from the target user's customized and private default information database, and directly returns the answer data to the user if it can be found. If the default information database cannot resolve the user's question, it utilizes the user's private knowledge base and large language model to obtain the answer data. This improves the efficiency of resolving user questions and saves system resources while ensuring the accuracy of the answer data. Furthermore, the default information database and / or knowledge base contain the user's private official document information, and security management ensures user data security. Moreover, this solution can be deployed on the user's internal system or a designated cloud server according to user needs, preventing the user's private data from being stolen externally, achieving closed and independent operation, and ensuring data security.

[0087] Figure 2 A flowchart illustrating a default information base construction method provided in an embodiment of this application is shown. This default information base construction method can be applied to document information security management methods based on large language models. Figure 2 As shown, the method includes the following steps:

[0088] Step S210: Obtain the data uploaded by the target user through the default information database configuration entry, and use the uploaded data as the default information database data.

[0089] This approach provides users with a visual entry point for configuring a default information repository, allowing target users to upload relevant documents and other data. Furthermore, it establishes a mapping between the uploaded data and the target user, using the uploaded data as the target user's default information repository data.

[0090] Step S220: Obtain the target user's historical browsing data and historical browsing frequency, and determine the default information database data from the historical browsing data based on the historical browsing frequency.

[0091] In this approach, the system also provides users with a management portal for relevant data within the system. Users with viewing permissions can view relevant data within the system through this management portal. During this process, the system records the target user's historical data access and the historical frequency of historical data access.

[0092] Furthermore, historical data with a frequency higher than a preset frequency or ranking in the top N (sorted from high to low) of historical browsing data are selected as the default information database data. This method allows for the selection of historical browsing data with high frequency from the target user as the default information database data, achieving automatic construction of the default information database and ensuring a better match between the constructed default information database and the corresponding target user. It should be understood that this invention obtains user-related information only with user authorization.

[0093] Step S230: Obtain the historical user question association data of the target user, and determine the default information database data from the association data based on the frequency of occurrence of the association data and the target user's response feedback.

[0094] In this approach, the system records the user questions provided by the target user in the past (i.e., the target user's historical user questions) and records the associated data of the historical user questions obtained from the knowledge base during the process of processing any historical user question.

[0095] For any associated data, the frequency of occurrence of associated data of historical user questions as the target user is statistically analyzed, and the feedback information of the target user on the response data obtained based on the associated data (i.e. the target user's response feedback) is obtained. The response feedback reflects the target user's satisfaction with the response data.

[0096] Furthermore, the default information base data is determined by combining the frequency of occurrence of related data and the response feedback. For example, if the target user's historical user questions used related data A in the knowledge base, and the frequency of occurrence of related data A in the processing of the target user's historical user questions is higher than a preset threshold or ranks in the top M (sorted from high to low), and the target user's response feedback to the response data based on the related data is positive feedback such as satisfaction, then related data A is used as the default information base data to improve the matching degree between the default information base and the target user.

[0097] Step S240: Store the default information database data in the default information database corresponding to the target user.

[0098] The default information database data will be determined and stored in the default information database corresponding to the target user. This data can be preprocessed before being stored in the default information database. This preprocessing includes, but is not limited to, removing invalid fields and format conversion (e.g., converting documents to Markdown or PDF formats).

[0099] Step S250: Based on the hit frequency and data storage duration of the default information database data, determine the data to be eliminated from the default information database; remove the eliminated data from the default information database and store the eliminated data in the knowledge base.

[0100] If a piece of data in the target user's default information database can be used as an answer to the user's question, it indicates that the data in that database matches the user's question. In this embodiment, the hit count of each piece of data in the default information database can be counted, and discarded data can be determined from the stored default information database data based on its data storage duration. For example, if the hit count of a piece of data is less than a preset number, or if its hit count is ranked X positions below the preset number, and its data storage duration exceeds a preset duration, then that piece of data is discarded and deleted from the default information database. Furthermore, the deleted default information database entries can be transferred to the knowledge base to facilitate the knowledge base's processing of user questions.

[0101] Therefore, the embodiments of the present invention can construct a default information database with a high degree of matching with the target user, thereby improving the hit rate of the default information database for the target user's user questions, improving the processing efficiency of user questions and saving system resources; moreover, timely updating of the default information database and recording the eliminated data in the default information database in the knowledge base improves data utilization.

[0102] Figure 3 A flowchart illustrating a method for retrieving related data according to an embodiment of this application is shown. This method can be applied to document information security management methods based on large language models. Figure 3 As shown, the method includes the following steps:

[0103] Step S310: Pre-build the knowledge base.

[0104] Specifically, the document is retrieved from the knowledge base, and preprocessed, such as by removing invalid fields and converting the format, while preserving the original format of the document.

[0105] Further, the knowledge base documents are segmented using a matching segmentation method based on their document type to obtain at least one knowledge base data. Specifically, for any given knowledge base document, its document type is determined. If the document is a long text type, natural language semantic segmentation can be used to ensure data language integrity; if the document is a structured data type, it is segmented according to the data structure (such as data records or fields) to preserve data integrity and consistency. The order of the segmented knowledge base data is also marked, and the knowledge base data and its corresponding order identifier are stored in the knowledge base.

[0106] In one optional implementation, during the segmentation process, the segmentation size range can be dynamically determined based on the number of tokens in the large language model. Segmentation is then performed according to this segmentation size range, ensuring that the number of tokens contained in each knowledge base data obtained from the segmentation falls within this segmentation size range. Different numbers of tokens in the large language model correspond to different segmentation size ranges.

[0107] In actual implementation, user questions provided by target users are obtained, and the default information database corresponding to the target user is determined; if there is no answer data matching the user question in the default information database, the subsequent steps are executed.

[0108] Step S320: Determine whether the similarity between the user's question and the knowledge base data is greater than the first similarity threshold; if yes, proceed to step S330; if no, proceed to step S340.

[0109] The similarity between user questions and segmented knowledge base data is calculated. Different processing methods are applied based on the similarity results. Specifically, both user questions and knowledge base data are vectorized, and the similarity between the user questions and knowledge base data is obtained based on the vector similarity.

[0110] In one optional implementation, to improve the efficiency of related data retrieval, candidate knowledge base documents can be identified first based on factors such as the similarity between the user question and the knowledge base document, the recall rate of the knowledge base document, and the credibility of the knowledge base document. Related data can then be searched preferentially from the knowledge base data segmented from these candidate documents. For example, if a knowledge base document has a high similarity between its document tags and the user question, a high historical recall rate, and a high historical credibility, then that knowledge base document is selected as a candidate knowledge base document, and the similarity between the user question and the knowledge base data in that candidate document is compared preferentially.

[0111] Step S330: Use the knowledge base data as associated data for the user's question.

[0112] If the similarity between knowledge base data and user questions is greater than the first similarity threshold, it indicates that the knowledge base data and user questions are extremely similar, and the knowledge base data is then used as the associated data for the user questions.

[0113] Step S340: Determine whether the similarity between the user's question and the knowledge base data is greater than the second similarity threshold; if yes, proceed to step S350; if no, proceed to step S360.

[0114] The second similarity threshold is less than the first similarity threshold, and the second similarity threshold is the minimum similarity threshold. If the similarity between the user's question and the knowledge base data is determined to be less than or equal to the first similarity threshold, it is further determined whether the similarity exceeds the minimum similarity threshold.

[0115] Step S350: Use the knowledge base data as candidate related data, and determine the related data of the user's question from the candidate related data and the context data of the candidate related data.

[0116] If the similarity between knowledge base data and user question is less than or equal to the first similarity threshold and greater than the second similarity threshold, then the knowledge base data is used as candidate related data. The related data of the user question is determined from the candidate related data and the context data of the candidate related data. That is, when the similarity is between the first similarity threshold and the second similarity threshold, a second segmentation is performed to accurately determine the related data.

[0117] In one optional implementation, for any candidate related data, context data of the candidate related data can be obtained according to the segmentation order of the candidate related data. The candidate related data and its context data are further segmented a second time, and the related data of the user question is determined based on the second segmentation result. Specifically, the candidate related data and its context data are cascaded segmented. The second segmentation result differs from the first segmentation result. The similarity between the user question and the knowledge base data after the second segmentation is further calculated. If the similarity is small compared to the first segmentation similarity, the knowledge base data with the highest similarity to the user question is extracted from the second segmentation knowledge base data as the related data. If the similarity is large compared to the first segmentation similarity, and the similarity is less than the first segmentation similarity, the second segmentation is performed again until the similarity after the second segmentation is higher than or close to the first segmentation similarity. The knowledge base data with the highest similarity to the user question is then extracted from the second segmentation knowledge base data as the related data.

[0118] Step S360: Determine that there is no related data for user questions in the knowledge base.

[0119] If the similarity between each knowledge base entry and the user's question is less than or equal to the second similarity threshold, then it is determined that there is no related data for the user's question in the knowledge base.

[0120] Therefore, the embodiments of the present invention use a matching method to determine the associated data of user questions based on the different similarities between user questions and knowledge base data, thereby improving the accuracy of determining associated data and further improving the processing accuracy of user questions.

[0121] Figure 4 This diagram illustrates the structure of a document information security management system based on a large language model, as provided in an embodiment of this application. Figure 4 As shown, the system 400 includes: an acquisition module 410, a default information base module 420, a knowledge base module 430, and a large language model module 440.

[0122] The acquisition module 410 is used to acquire user questions provided by the target user and determine the default information database corresponding to the target user.

[0123] The default information base module 420 is used to send the response data from the default information base to the target user if there is response data in the default information base that matches the user's question.

[0124] The knowledge base module 430 is used to determine whether there is related data for the user's question in the knowledge base if there is no answer data matching the user's question in the default information base.

[0125] The large language model module 440 is used to input the associated data and user questions into the large language model if the condition is met, and then feed back the response data output by the large language model to the target user if the condition is not met.

[0126] The default information base and / or the knowledge base are managed using a preset management algorithm for security purposes, and the default information base and / or the knowledge base contain at least one official document.

[0127] In one optional implementation, the default information base module is used to: obtain data uploaded by the target user through the default information base configuration entry, and use the uploaded data as the default information base data;

[0128] And / or, obtain the target user's historical browsing data and historical browsing frequency, and determine the default information database data from the historical browsing data based on the historical browsing frequency;

[0129] And / or, obtain the historical user question association data of the target user, and determine the default information database data from the association data based on the frequency of occurrence of the association data and the target user's response feedback;

[0130] The default information database data is stored in the default information database corresponding to the target user.

[0131] In one optional implementation, the default information base module is used to: determine the discarded data from the default information base based on the hit frequency of the default information base data and the data storage duration;

[0132] The knowledge base module is used to remove obsolete data from the default information database and store obsolete data in the knowledge base.

[0133] In one alternative implementation, the knowledge base module is used to: retrieve knowledge base documents;

[0134] The knowledge base documents are segmented according to their document types using a matching segmentation method to obtain at least one knowledge base data.

[0135] Store knowledge base data in the knowledge base.

[0136] In one optional implementation, the knowledge base module is used to: if the similarity between knowledge base data and user question is greater than a first similarity threshold, then use the knowledge base data as associated data of the user question;

[0137] If the similarity between knowledge base data and user question is less than or equal to the first similarity threshold and greater than the second similarity threshold, then the knowledge base data is used as candidate related data, and the related data of user question is determined from the candidate related data and the context data of the candidate related data.

[0138] If the similarity between each knowledge base entry and the user's question is less than or equal to the second similarity threshold, then it is determined that there is no related data for the user's question in the knowledge base.

[0139] The second similarity threshold is less than the first similarity threshold.

[0140] In one optional implementation, the knowledge base module is used to: for any candidate related data, perform a secondary segmentation on the candidate related data and its context data, and determine the related data of the user's question based on the secondary segmentation result.

[0141] Therefore, this embodiment of the invention prioritizes searching for answer data for user questions from the target user's customized and private default information database, and directly returns the answer data to the user if it can be found. If the default information database cannot resolve the user's question, it utilizes the user's private knowledge base and large language model to obtain the answer data. This improves the efficiency of resolving user questions and saves system resources while ensuring the accuracy of the answer data. Furthermore, the default information database and / or knowledge base contain the user's private official document information, and security management ensures user data security. Moreover, this solution can be deployed on the user's internal system or a designated cloud server according to user needs, preventing the user's private data from being stolen externally, achieving closed and independent operation, and ensuring data security.

[0142] This application provides a non-volatile computer storage medium storing at least one executable instruction or computer program that enables a processor to perform the operation corresponding to the document information security management method based on a large language model in any of the above method embodiments.

[0143] This application provides a computer program product, which includes at least one executable instruction or computer program that enables a processor to perform the operation corresponding to the document information security management method based on a large language model in any of the above method embodiments.

[0144] Figure 5 The diagram shows a structural schematic of a computing device provided in an embodiment of this application. The specific embodiments of this application do not limit the specific implementation of the computing device.

[0145] like Figure 5 As shown, the computing device may include: a processor 502, a communications interface 504, a memory 506, and a communications bus 508.

[0146] The processor 502, communication interface 504, and memory 506 communicate with each other via communication bus 508. Communication interface 504 is used to communicate with other network elements such as clients or other servers. The processor 502 executes program 510, specifically performing the relevant steps in the above-described embodiment of the document information security management method based on a large language model for computing devices.

[0147] Specifically, program 510 may include program code that includes computer operation instructions.

[0148] Processor 502 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The computing device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0149] Memory 506 is used to store program 510. Memory 506 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device. Program 510 can specifically be used to cause processor 502 to perform the operations described in the method embodiments above.

[0150] In summary, based on the computing device, computer storage medium, and computer program products provided in this embodiment, the system prioritizes searching for answer data for user questions from the target user's customized and private default information database. If the answer is found, it is directly returned to the user. If the default information database cannot resolve the user's question, the system utilizes the user's private knowledge base and large language model to obtain the answer data. This improves the efficiency of resolving user questions while ensuring the accuracy of the answer data and saving system resources. Furthermore, the default information database and / or knowledge base contain the user's private official document information, and security management ensures user data security. Moreover, this solution can be deployed on the user's internal system or a designated cloud server according to user needs, preventing the theft of user's private data from external sources, achieving closed and independent operation, and ensuring data security.

[0151] The algorithms or displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of this application are not directed to any particular programming language. It should be understood that the content of this application described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of this application.

[0152] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0153] Similarly, it should be understood that, in order to simplify this application and aid in understanding one or more of the various inventive aspects, features of the embodiments of this application are sometimes grouped together in a single embodiment, figure, or description thereof in the above description of exemplary embodiments of this application. However, this method of disclosure should not be construed as reflecting an intention that the claimed application requires more features than expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0154] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0155] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, in the following claims, any of the claimed embodiments can be used in any combination.

[0156] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of this application. This application can also be implemented as a device or system program (e.g., a computer program and computer program product) for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0157] It should be noted that the above embodiments are illustrative of this application and not restrictive, and those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several systems, several of these systems may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method for managing the security of official document information based on a large language model, characterized in that, include: Obtain user questions provided by the target users and determine the default information database corresponding to the target users; If there is a matching answer in the default information database, the answer in the default information database will be sent back to the target user. If no answer data matching the user's question exists in the default information base, determine whether there is related data for the user's question in the knowledge base; If so, input the associated data and user questions into the large language model, and feed back the response data output by the large language model to the target user; If not, input the user's question into the large language model and feed back the response data output by the large language model to the target user; The default information base and / or the knowledge base are managed using a preset management algorithm for security purposes, and the default information base and / or the knowledge base contain at least one official document.

2. The method according to claim 1, characterized in that, The default information base is constructed in the following manner: Obtain the data uploaded by the target user through the default information database configuration entry, and use the uploaded data as the default information database data; And / or, obtain the target user's historical browsing data and historical browsing frequency, and determine the default information database data from the historical browsing data based on the historical browsing frequency; And / or, obtain the historical user question association data of the target user, and determine the default information database data from the association data based on the frequency of occurrence of the association data and the target user's response feedback; The default information database data is stored in the default information database corresponding to the target user.

3. The method according to claim 2, characterized in that, The method further includes: Based on the hit frequency and data storage duration of the default information database, the data to be eliminated is determined from the default information database. Remove obsolete data from the default information base and store the obsolete data in the knowledge base.

4. The method according to any one of claims 1-3, characterized in that, The method further includes: Retrieve knowledge base documents; The knowledge base documents are segmented according to their document types using a matching segmentation method to obtain at least one knowledge base data. Store knowledge base data in the knowledge base.

5. The method according to claim 4, characterized in that, The determination of whether the knowledge base contains related data for user questions includes: If the similarity between knowledge base data and user question is greater than the first similarity threshold, then the knowledge base data will be used as the associated data of the user question. If the similarity between knowledge base data and user question is less than or equal to the first similarity threshold and greater than the second similarity threshold, then the knowledge base data is used as candidate related data, and the related data of user question is determined from the candidate related data and the context data of the candidate related data. If the similarity between each knowledge base entry and the user's question is less than or equal to the second similarity threshold, then it is determined that there is no related data for the user's question in the knowledge base. The second similarity threshold is less than the first similarity threshold.

6. The method according to claim 5, characterized in that, The process of determining the association data of the user question from the candidate association data and the context data of the candidate association data includes: For any candidate related data, the candidate related data and its context data are further segmented, and the related data of the user's question are determined based on the results of the secondary segmentation.

7. A document information security management system based on a large language model, characterized in that, include: The acquisition module is used to acquire user questions provided by the target user and determine the default information database corresponding to the target user. The default information database module is used to send the answer data from the default information database to the target user if there is answer data in the default information database that matches the user's question. The knowledge base module is used to determine whether there is related data for the user's question in the knowledge base if no answer data matching the user's question exists in the default information base. The large language model module is used to input the relevant data and user questions into the large language model if the condition is met, and then return the response data output by the large language model to the target user if the condition is not met. The default information base and / or the knowledge base are managed using a preset management algorithm for security purposes, and the default information base and / or the knowledge base contain at least one official document.

8. A computing device, characterized in that, include: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the document information security management method based on a large language model as described in any one of claims 1-6.

9. A computer storage medium, characterized in that, The storage medium stores at least one executable instruction, which causes the processor to perform the operation corresponding to the document information security management method based on a large language model as described in any one of claims 1-6.

10. A computer program product, characterized in that, It includes at least one executable instruction that causes the processor to perform the operation corresponding to the document information security management method based on a large language model as described in any one of claims 1-6.

Images