Intelligent voice interaction official document exchange system based on large model and man-machine conversation control method

Abstract

The invention relates to the technical field of official document exchange, and discloses an intelligent voice interaction official document exchange system based on a large model and a man-machine conversation control method, and the system comprises a confidential document management subsystem which is used for the transmitting and receiving business management of a document; the voice interaction module is used for receiving and analyzing a user voice instruction; the control module is used for scheduling each sub-module in the system to complete an operation instruction; the data processing module is used for analyzing and processing contents, states and circulation of documents and letters based on a large model; and the feedback module is used for feeding back the execution result to the user in a voice or text form. According to the technical scheme, a three-member management platform is adopted to perform labor division management and real-time safety auditing on the system permission, the technical effect of improving the system permission distribution reasonability and the overall operation safety is achieved, and compared with the technical scheme of concentrated permission and lack of effective auditing and safety management in the prior art, the technical scheme has the advantages that the system permission distribution reasonability and the overall operation safety are improved. And the defects of permission abuse and security vulnerability hidden dangers are overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of official document exchange, and in particular to an intelligent voice interactive official document exchange system based on a large model and a human-computer dialogue control method. Background Art

[0002] With the continuous improvement of informatization level, official document exchange systems have been widely used in office scenarios in government agencies, enterprises, institutions and other fields. Official document exchange systems are mainly used to handle processes such as document transmission, instructions and archiving between institutions, and usually need to have efficient, secure and controllable characteristics. In traditional official document management, document transmission relies on paper media or simple e-mail forms, and there are problems such as low transmission efficiency and difficulty in process tracking. Modern official document exchange systems use digital means to realize electronic transmission and partial automation of documents, and gradually replace traditional manual operations. However, in the face of increasingly complex application requirements and highly sensitive official document management scenarios, the existing official document exchange systems still have obvious deficiencies in intelligence, security and adaptation to the local environment, and it is difficult to meet the office requirements of the new era.

[0003] Currently, many official document exchange systems rely on foreign technology or closed proprietary architectures, resulting in many problems in practical applications. On the one hand, the advancement of the localization process requires that the official document exchange system be fully adaptable to the domestic environment in terms of hardware and software architecture, including support for processors, operating systems and databases. On the other hand, existing systems often lack sufficient intelligent processing capabilities, especially in key links such as authority management, official document classification, flow monitoring and emergency task reminders. Most of them rely on manual intervention, which is inefficient and error-prone. In addition, the potential information security risks brought about by the reliance on foreign technology, especially in confidential document management scenarios, may lead to the leakage or tampering of sensitive data. Therefore, it is imperative to design a document exchange system based on a domestic platform that can achieve intelligent management and has high security.

[0004] The existing official document exchange system mainly has the following problems and shortcomings: First, in terms of authority management, the traditional system has centralized authority configuration, and the risk of authority abuse and unauthorized operation is high. At the same time, it lacks real-time auditing functions and insufficient compliance monitoring of system operations. Secondly, in the processing of official document content, the existing system usually adopts fixed rules for classification and flow status tracking. This method has poor adaptability to complex scenarios and is prone to processing errors or delays due to insufficient rules. In addition, in the management of urgent official documents, most existing technologies rely on manual marking of urgent tasks, lack real-time monitoring and multi-channel reminder functions, are prone to response delays, and cannot meet the needs of efficient processing. Finally, the existing system has a low degree of intelligence, and the modular design is severely fragmented, making it difficult to achieve intelligent collaboration throughout the entire process, especially in a domestic environment. Its functional scalability and compatibility are insufficient. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides an intelligent voice interactive document exchange system and a human-computer dialogue control method based on a large model, which solves the problem that the document exchange system mostly relies on manual intervention, is inefficient and prone to errors in key links such as authority management, document classification, flow monitoring and emergency task reminders.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an intelligent voice interactive document exchange system and a human-computer dialogue control method based on a large model, comprising: Confidential document management subsystem, used for document sending and receiving business management; Voice interaction module, used to receive and analyze user voice commands; The control module is used to schedule the sub-modules in the system to complete the operation instructions; The data processing module analyzes and processes the content, status and flow of official documents and letters based on the big model; The feedback module is used to feed back the execution results to the user in the form of voice or text.

[0007] Preferably, the confidential document management subsystem performs automatic classification and semantic analysis of document content based on a large model, and realizes intelligent management through keyword extraction and context analysis of text content. The large model is a pre-trained language model built based on deep learning technology. Through training on massive corpus, it has powerful semantic understanding and context analysis capabilities, and can perform accurate classification and semantic analysis on document content.

[0008] Preferably, the voice interaction module uses a pre-trained language model to perform intent recognition and context analysis on the user's voice, generates operation instructions corresponding to the voice content, and schedules the corresponding functional modules to execute through the control module.

[0009] Preferably, the feedback module includes: A feedback generation unit, used for converting the execution result into a voice or graphical interface; The prompt unit is used to attract the user's attention through alarms or highlighted prompts when urgent items are being processed.

[0010] Preferably, the data processing module extracts document semantic information through a pre-trained language model, and automatically classifies, sorts, and tracks the status of official documents in combination with the structural features of the official documents.

[0011] Preferably, the system further comprises: Self-service platform system, installed in the exchange room, with functions including mail registration and barcode printing for confidential mail receiving and dispatching personnel; The department inquiry system is used by confidential mail clerks in networked departments to conduct inquiry and statistical management of the department's official documents and letters.

[0012] Preferably, the system further comprises: Urgent reminder system, including: Urgent item detection unit, used to monitor the status of urgent items waiting to be picked up in the exchange box; A reminder trigger unit, for generating a reminder signal after detecting an urgent item; A reminder display unit, used to display reminder information on a user terminal or a self-service device in an exchange room; The three-member management platform includes: System administrator module, used to allocate and manage system basic data and user resource permissions; Audit administrator module, used to record and analyze system operation logs and check for violations; The security administrator module is used to review the permission resources allocated by the system administrator and ensure the security of user permissions.

[0013] A human-computer dialogue control method for an intelligent voice interactive document exchange system based on a large model comprises the following steps: The user inputs commands via voice; The system uses the voice interaction module to analyze the voice content and generate operation instructions based on the big model; The control module schedules the function modules in the system to execute the instructions; The feedback module outputs the operation results to the user in the form of voice or text.

[0014] Preferably, the analysis step includes: using a large model to deeply analyze the semantics of the user's voice content, and generating accurate operation instructions in combination with the user's historical instructions and context information.

[0015] Preferably, the feedback step includes: Generate text content for the execution results and generate voice feedback through the speech synthesis module; The execution status is displayed on the touch screen of the self-service platform in a graphical interface for users to review.

[0016] The present invention provides an intelligent voice interactive document exchange system and a human-computer dialogue control method based on a large model. It has the following beneficial effects: 1. The present invention achieves the technical effect of improving the rationality of system authority allocation and overall operational security by adopting a three-member management platform to carry out division of labor management and real-time security audit of system authority. Compared with the technical solution of centralized authority and lack of effective audit and security management in the prior art, it solves the problems of authority abuse and security vulnerabilities, ensures the standardization and security of the official document management system, and provides strong support for the use of highly sensitive scenarios.

[0017] 2. The technical solution of the present invention uses a data processing module to perform semantic extraction, classification and circulation status tracking on the content of official documents, achieving the technical effect of intelligent and automated optimization of the official document management process. Compared with the technical solution in the prior art that relies on manual classification and has low processing efficiency, it solves the shortcomings of cumbersome manual operations, low classification efficiency and delayed status updates in traditional official document management, and significantly improves the operating efficiency and accuracy of the official document management system.

[0018] 3. The present invention realizes real-time monitoring and multi-form reminder functions of emergency documents by designing an urgent document reminder system, achieving the technical effect of rapid response and efficient processing of emergency documents. Compared with the technical solution in the prior art that emergency documents rely on manual recognition and are prone to delays, it solves the problem of untimely processing of emergency tasks in traditional document management and ensures the timeliness and reliability of document circulation.

[0019] 4. The present invention uses a data processing module to perform semantic extraction, classification and circulation status tracking on the official document content, achieving the technical effect of intelligent and automated optimization of the official document management process. Compared with the technical solutions in the prior art that rely on manual classification and have low processing efficiency, it solves the shortcomings of cumbersome manual operations, low classification efficiency and delayed status updates in traditional official document management, and significantly improves the operating efficiency and accuracy of the official document management system. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a system framework diagram of the present invention; Figure 2 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] Please see attached Figure 1,The embodiment of the present invention provides an intelligent voice interactive document exchange system based on a large model, including: a confidential document management subsystem for document sending and receiving business management; Voice interaction module, used to receive and analyze user voice commands; The control module is used to schedule the sub-modules in the system to complete the operation instructions; The data processing module analyzes and processes the content, status and flow of official documents and letters based on the big model; The feedback module is used to feed back the execution results to the user in the form of voice or text.

[0023] Specific implementation method of confidential document management subsystem The intelligent voice interaction official document exchange system based on the big model of the present invention aims to realize the automation and intelligent operation of the whole process of sending, receiving, managing, and counting official documents through the powerful semantic understanding and voice processing capabilities of the big model. As a core functional module, the confidential document management subsystem works in coordination with the voice interaction module, the data processing module, and the control module to jointly complete the efficient and secure management of confidential documents. This embodiment specifically describes in detail the functional implementation, working principle, and related technical details of the confidential document management subsystem to ensure that those skilled in the art can reproduce the technology based on the disclosed content.

[0024] In this embodiment, the core of the confidential document management subsystem is to achieve full-process management of confidential documents from sending and receiving to archiving through the deep semantic analysis and natural language processing capabilities of the large model. The system receives user voice commands through linkage with the voice interaction module, and uses the data processing module to perform semantic analysis and classification processing on the file content; at the same time, the control module schedules the collaboration of other subsystems to complete the status tracking and feedback of the file.

[0025] In some embodiments, the functions of the confidential document management subsystem include but are not limited to document receipt registration, document issuance registration, document classification, instruction management, document query, document statistics, pending document management and data export, etc. In general, after receiving user instructions through the voice interaction module, the system interprets them into operation instructions, and then the control module calls the corresponding function to complete the specific operation.

[0026] In this embodiment, the document registration function uses the natural language processing capabilities of the large model to realize the automatic collection and classification of official documents. Specifically, when the user inputs the command "register document receipt" through the voice interaction module, the system uses the pre-trained language model to perform semantic analysis on the voice content and extract key elements such as the file title, source unit, date, etc. Through the context analysis of the text content, the system can automatically identify the urgency and category of the file. As an option, the system can also automatically specify a handler or department for the document according to the user's authority and usage scenario.

[0027] In a possible implementation, the processing of receiving document registration can be described by the following formula: S=f(T,K,C) Among them, S represents the document registration result, T is the main content of the file, K is the extracted keyword, C is the file category weight calculated by the large model through context analysis, and function f is the semantic parsing and classification function implemented by the large model. By adjusting the weights of parameters K and C, the system can optimize the classification results of files according to different scenarios.

[0028] In this embodiment, the document issuance registration function is similar to the document receipt registration, but adds a marking function for the file flow path and processing priority. Specifically, after the user inputs the document content through voice or text, the system uses the data processing module to perform semantic extraction on the content, and automatically generates a document issuance flow plan through the control module. As an option, the system can also prompt the user through alarms or highlights in the feedback module according to the processing requirements of urgent files.

[0029] The instruction management function relies on the semantic understanding ability of the big model to realize the automatic classification and tracking of document approval opinions. In some embodiments, the system automatically classifies the document into different processing states by extracting the instruction information in the document content (such as "agree", "under review", "need to be modified", etc.), and generates corresponding task reminders. As a possible implementation method, the system describes the analysis of instruction content through the following formula: B=Extract(T,Keywords) Among them, B is the content of the instruction information, T is the content of the document body, Keywords is a set of predefined keywords related to the instruction, and the system extracts and stores the instruction letter through keyword matching and context semantic analysis.

[0030] In the file query function, users can input query conditions through voice or text. The system uses the big model for semantic matching and automatically retrieves file records that meet the conditions. For example, when the user enters "query all posts this month", the system will extract relevant file records based on timestamps and keywords, and return the query results in the form of voice or text through the feedback module.

[0031] The file statistics function analyzes all the official document data in the system to achieve statistics and visualization of information such as the amount of sending and receiving, type distribution, etc. In some embodiments, the system uses a large model to perform semantic analysis on the file title and text content, extracts keywords and classification information, and generates statistical reports. In one possible implementation, the generation of statistical data can be described by the following formula: Among them, R represents the statistical result, T iand C i are the text content and classification weight of the i-th file respectively, and N is the total number of files. By summarizing the classification weights of all files, the system can generate statistical data on the number and type of files.

[0032] The file statistics function analyzes all the official document data in the system to achieve statistics and visualization of information such as the amount of sending and receiving, type distribution, etc. In some embodiments, the system uses a large model to perform semantic analysis on the file title and text content, extracts keywords and classification information, and generates statistical reports. In a possible implementation, the generation of statistical data can be described by the following formula: Among them, represents the statistical result, and are the text content and classification weight of the ith file, respectively, and is the total number of files. By summarizing the classification weights of all files, the system can generate statistical data on the number and type of files.

[0033] Specific implementation method of voice interaction module The voice interaction module of the intelligent voice interaction official document exchange system based on the big model of the present invention is the core module for voice information interaction between the system and the user, and its main function is to receive user voice commands, parse them and generate executable operation commands. The module realizes a full-link closed loop from voice input to system command generation through collaborative work with the control module, the confidential document management subsystem and the data processing module, providing technical support for the automation and intelligence of official document exchange. The voice interaction module relies on the deep semantic understanding and context analysis capabilities of the big model, and shows high accuracy and robustness in voice-to-text, intent recognition and command generation.

[0034] In this embodiment, the main workflow of the voice interaction module includes the steps of voice reception, voice-to-text conversion, semantic analysis, and operation instruction generation. Generally, the user inputs the demand through voice, such as "query unprocessed files" or "register and issue a document", and the voice interaction module converts the voice signal into a text representation that can be understood by the system, and then uses the big model to deeply analyze the text content, identify the user's intention, generate corresponding operation instructions, and schedule the relevant functional modules to execute through the control module.

[0035] In some embodiments, the voice interaction module implements deep semantic analysis of voice content based on a pre-trained language model (such as a GPT series model). Specifically, the module adopts the following processing logic: First, the voice signal is collected by the voice receiving unit, and the voice signal is converted into text data T using a voice recognition algorithm. During the voice recognition process, the system adopts an algorithm based on a combination of an acoustic model and a language model to improve the recognition accuracy of the voice content, especially in noisy environments or voice scenes with complex accents. Secondly, the converted text data T is input into the semantic parsing submodule, and the natural language processing capabilities of the large model are used to perform word segmentation, part-of-speech tagging, and context analysis on the text content, thereby extracting key semantic elements in the user's instructions. For example, when inputting "query for this month's receipt", the large model can identify the keywords "query", "this month", and "receive" through context analysis, and generate the following operation intentions: I=f(T,C) Among them, I is the generated operation intention, T is the text input, C is the context information (such as the user's historical operation records and the current module status), and the function f represents the semantic parsing function implemented by the large model, which optimizes the matching of semantic elements in combination with user behavior data.

[0036] In one possible implementation, in order to improve the accuracy and robustness of parsing, the system adopts a semantic parsing model based on a multi-layer attention mechanism to perform layer-by-layer semantic abstraction of the text content and generate accurate operation intentions in combination with contextual information. Specifically, the attention mechanism can identify the most important words or phrases in the text by calculating the semantic weight matrix W: Among them, Q and K are query and key value vectors, d k is the vector dimension, and W is the weight matrix, which indicates the relevance of the word to the overall semantics.

[0037] After the operation intention is generated, the voice interaction module converts it into an operation instruction O that can be executed by the system. The generation process of the operation instruction depends on the logical reasoning and intention matching capabilities of the big model for the text content. For example, when the user enters "register to post a document", the system generates the following operation instruction by analyzing the subject of the post, the receiving unit and other elements: O={A,P,D} Among them, A is the action type (such as "register"), P is the relevant parameters (such as file title, receiving unit, etc.), and D is the target data of the operation.

[0038] In some embodiments, the voice interaction module also combines the user's historical interaction records and uses the context memory capability of the large model to optimize the operation instruction generation process. For example, when a user queries similar content multiple times (such as "query unprocessed files"), the system can automatically identify the user's behavior habits and generate operation suggestions in advance. This function is implemented by the following formula: O=g(I,H) Among them, H is the user's historical interaction record, g is the operation optimization function, and the optimized operation instructions are generated by combining the historical data H and the current intention I.

[0039] The feedback mechanism is an important part of the voice interaction module. Generally, after the operation instruction is generated, the system will feedback the execution result to the user in the form of voice or text through the feedback module. In some embodiments, the voice interaction module also supports multiple rounds of dialogue. That is, when the user's needs are unclear, the system can further confirm the user's intention by asking voice questions. For example, when the user enters "query file", the system can ask "Are you querying for receiving or sending documents?" to ensure that the generated operation instruction has higher accuracy.

[0040] Specifically, the connection between the voice interaction module and the confidential document management subsystem is reflected in the command transmission and function call. For example, after the user issues the "register received document" command, the voice interaction module passes the generated operation instruction to the confidential document management subsystem through the control module, together with the necessary parameters (such as file title, source unit, etc.). Subsequently, the confidential document management subsystem executes the document receipt registration operation according to the instruction and returns the result to the voice interaction module, which then feeds back the result to the user.

[0041] Through the above process, the voice interaction module effectively connects user input with system function execution, making the entire document exchange system realize efficient and convenient human-computer interaction. Its deep semantic analysis capability based on large models not only improves the accuracy of voice command recognition, but also significantly reduces the threshold for users to learn system operation, thus providing technical support for the intelligent development of document management.

[0042] Specific implementation of data processing module In the intelligent voice interaction official document exchange system based on the big model of the present invention, the data processing module, as one of the core technical modules, is mainly responsible for analyzing and processing the official documents and correspondence data involved in the system. This module is based on the powerful natural language processing (NLP) capability and structured data processing capability of the big model, combined with the format characteristics and circulation status of the official document, to complete the functions of semantic analysis, automatic classification, status tracking and circulation path optimization of the official document content. The data processing module works in coordination with the voice interaction module and the control module to perform in-depth analysis on the command content input by the user to ensure that the generated operation results are accurate, and at the same time provide structured data support for the feedback module to ensure that the user can intuitively understand the operation results and their background information.

[0043] In this embodiment, the implementation of the data processing module is based on the comprehensive processing capabilities of the big model for text data and structured data. Generally, the data processing module receives the results of user intent analysis from the voice interaction module or the content of official documents received from the confidential document management subsystem, and performs in-depth processing on these data. In some embodiments, the data processing module further combines the user's historical operation records and contextual information to perform intelligent processing on the classification, priority evaluation and flow path optimization of official documents.

[0044] In this embodiment, the semantic parsing function of the data processing module is specifically implemented as follows: by using the big model to perform semantic analysis on the content of the official document, the system can extract key information elements in the document, including the subject, keywords, source unit and processing requirements, etc. For example, the official document content T uploaded or input by the user is input into the big model, and after word vector embedding and context semantic analysis, the key information set K = {k1, k2, ..., k n}, in one possible implementation, the key steps of semantic parsing include the following formula: in, is the keyword candidate set, v i is the word vector representation of each word in the text, α i is the context weight coefficient Sim(v i , k) is the similarity score between word vectors. Through the above formula, the system can automatically extract the keyword set K that best matches the semantic content from a large-scale corpus, and then further classify and process the official documents.

[0045] In this embodiment, the semantic parsing function of the data processing module is specifically implemented as follows: by using the big model to perform semantic analysis on the content of the official document, the system can extract key information elements in the document, including the subject, keywords, source unit and processing requirements, etc. For example, the official document content T uploaded or input by the user is input into the big model, and after word vector embedding and context semantic analysis, the key information set K = {k1, k2, ..., k n}, in one possible implementation, the key steps of semantic parsing include the following formula: in, is the keyword candidate set, v i is the word vector representation of each word in the text, α i is the context weight coefficient Sim(v i , k) is the similarity score between word vectors. Through the above formula, the system can automatically extract the keyword set K that best matches the semantic content from a large-scale corpus, and then further classify and process the official documents.

[0046] In this embodiment, the automatic classification function of the data processing module is specifically implemented as follows: Based on the structural features and semantic features of the document content, the system intelligently classifies the document. In some embodiments, the system constructs a classification model C by jointly analyzing the document body T and the title H. Specifically, the system first performs word vectorization on T and H and inputs them into the classification model C = {c1, c2, ..., c m}, generate the probability distribution of each classification label. The design formula of the classification model is as follows: P(c|T,H)=softmax(W·concat(E T ,E H )+b) Among them, E T and E H are the word vector embeddings of the text and title respectively, W is the weight matrix, b is the bias vector, and softmax is used to map the results into probability distribution. In a possible implementation, the system verifies the document classification results and adjusts the weights of the classification model based on the actual operation process to improve the classification accuracy. For example, the system will prioritize the classification of "urgent documents" to ensure that such documents are given priority in the circulation process.

[0047] In one possible implementation, the system verifies the results of document classification and adjusts the weight of the classification model based on the actual operating process to improve the accuracy of classification. For example, the system will prioritize the classification of "urgent documents" to ensure that such documents are given priority during the circulation process.

[0048] In this embodiment, the status tracking function of the data processing module is specifically implemented as follows: by monitoring the circulation path and processing status of the official document, the system can automatically update the status information of the official document and generate corresponding processing suggestions. In some embodiments, the data processing module combines the analysis of the official document content and circulation nodes with the large model to calculate the processing priority and target node of each file. Specifically, the priority evaluation formula of the file is as follows: Among them, P priority is the priority score, S i is the semantic feature score associated with the document (such as the number of occurrences of the keyword “urgent”), β i is the weight coefficient, which indicates the influence of different features on the priority. Through this formula, the system can dynamically adjust the flow path and nodes of the file.

[0049] In this embodiment, the data processing module also supports the visual display of file status. As an option, the system displays the flow status of files from "registration" to "instruction" to "archiving" through a graphical interface, and uses different colors or icons to identify the current status and processing nodes. For example, a file marked in red may indicate a "pending" status, while a green mark indicates an "archived" status. This function works in conjunction with the feedback module to provide users with intuitive feedback on operation results.

[0050] In this embodiment, the circulation path optimization function of the data processing module is specifically implemented as follows: combined with the prediction ability of the large model, the system calculates the best circulation path for the document. For example, for an urgent document, the system can give priority to pushing it to the relevant leader for approval, while ordinary documents are circulated according to the regular process. The goal of the circulation path optimization is to minimize the processing time T p , the formula is as follows: Among them, w i is the workload weight of the current node, v i The system dynamically adjusts the path by monitoring the working status of each node in real time to ensure efficient flow.

[0051] Through the implementation of the above functions, the data processing module can effectively improve the intelligence level of the official document management system. Its deep semantic parsing and structured data analysis capabilities based on large models ensure that the entire process of document processing from receipt to archiving is efficient, accurate and traceable, providing technical support for the intelligent operation of the official document exchange system.

[0052] Specific implementation methods of the self-service platform system and department query system The self-service platform system and department query system in the intelligent voice interactive official document exchange system based on the large model of the present invention are key modules that provide convenient user interaction and efficient query functions for official document exchange and management. These two modules respectively undertake the auxiliary operation of official document circulation and the query and statistical functions of department-level management in the overall system. The self-service platform system is directly associated with the confidential document management subsystem, which is used to realize the functions of document sending and receiving registration, barcode generation and printing, and provide users with intuitive interface interaction support; the department query system provides rapid retrieval and management decision support for department confidential personnel through comprehensive query and statistics of official document data within the department. These two modules are closely connected with the control module, voice interaction module and data processing module, and jointly construct an efficient process from operation to query of the system.

[0053] In this embodiment, the main functions of the self-service platform system include document sending and receiving registration, barcode printing, file status feedback and urgent document reminders, etc., aiming to provide convenient and efficient document processing support for confidential sending and receiving personnel. Generally, the self-service platform system is installed in the exchange room, and user operations are realized through touch screen or voice interaction. In some embodiments, the system realizes the user's document registration operation and generates a document barcode for printing by calling the function module of the confidential document management subsystem. The barcode information contains the unique identifier and flow information of the document.

[0054] Specifically, when a user initiates a post registration request on the self-service platform, the system first analyzes the user's voice or text input through the voice interaction module and converts the input content into structured data D = {d1, d2, ..., d n}, where d i Indicates the specific attributes of the document (such as title, recipient unit, urgency, etc.). Subsequently, the data processing module performs semantic analysis on D and generates a document barcode based on the context information. The barcode generation algorithm can be expressed as: B=f(D,M) Among them, B is the generated barcode, D is the input data set, and M is the document classification model. By parsing and encoding D, the barcode content is mapped to the corresponding classification and circulation path.

[0055] As an option, the self-service platform system also supports batch operations on multiple files. For example, when a user needs to register multiple documents in batches, the system automatically extracts the core content of each document through the batch data parsing function and generates barcodes one by one, thereby simplifying the user operation process. In one possible implementation, the self-service platform system automatically marks urgent documents and triggers a reminder function by embedding the priority judgment function of the large model. For example, when the keyword "urgent" is detected, the system will generate a high priority mark and remind the confidential sending and receiving personnel to handle it in time through highlighting or sound prompts. The priority score of the urgent document mark can be expressed as: Among them, P is the priority score, α i is the weight of the keyword, w i is the frequency of keyword occurrence. The system adjusts the weight coefficient α in real time. i To optimize the accuracy of priority scoring.

[0056] In this embodiment, the function of the department query system is mainly reflected in the rapid retrieval and statistical management of the sending and receiving of official documents. Generally, department confidential personnel access the system through a networked query terminal to query and statistically analyze the sending and receiving records of the department's official documents and letters. In some embodiments, the department query system deeply analyzes the query request and quickly returns the result by calling the analysis capability of the data processing module.

[0057] Specifically, when the user enters the query conditions (such as time range, document category, issuing unit, etc.), the system first receives the user request through the voice interaction module or text input module and generates a query instruction Q. The construction logic of the query instruction is as follows: Q={f1,f2,...,f m} Among them, f i Represents each query condition (such as time range or category filter condition). Subsequently, the system uses the data processing module to perform semantic matching on the query condition and retrieve a set of records R that meet the conditions. The time it takes to return the query result depends on the degree of optimization of the retrieval algorithm and the indexing strategy of the database.

[0058] As an option, the department query system also supports the automatic generation of statistical charts. By calling the statistical analysis submodule in the data processing module, the system can classify and count the query results and generate visual charts. For example, when the user queries "this month's received document statistics", the system will calculate the number of received documents based on the timestamp and classification label of the received documents and generate a bar chart or pie chart. The calculation formula for statistical data is as follows: Among them, S is the statistical result, C is the classification set, count(c,T) represents the number of files of classification c within the time range T. In a possible implementation, the department query system further combines the processing status data of the official documents to filter and highlight the unprocessed or overdue files. For example, when the system detects that some files exceed the specified processing time, it will automatically mark these files and issue a prompt to the user. The filtering rules for overdue files can be expressed as: F={f|t f >t d} Among them, F is the set of overdue files, t f is the actual processing time of the file, t d Provide for the processing time of the document.

[0059] Through the above functions, the self-service platform system and the department query system play an important role in the circulation and management of official documents. The self-service platform system realizes the efficient registration and barcode generation of official document receipt and delivery, while the department query system provides strong support for department-level official document management through rapid retrieval and statistical analysis. The coordinated operation of the two not only improves the intelligence level of the entire system, but also provides users with a convenient and efficient interactive experience, thereby fully meeting the needs of confidential document management.

[0060] Specific implementation method of the urgent reminder system The urgent reminder system in the intelligent voice interactive document exchange system based on the large model of the present invention is a functional module designed to ensure the timely processing of urgent documents. This module automatically identifies the files that need to be processed urgently by real-time monitoring of the status and priority of the documents, combined with the content characteristics and circulation requirements of the urgent documents, and promptly issues reminders to the user. The urgent reminder system works closely with the data processing module, the self-service platform system and the voice interaction module to jointly achieve accurate detection of urgent documents, status updates and multi-form reminder feedback. Its design goal is to improve the efficiency of urgent document processing and reduce the risk of delays, thereby ensuring the circulation speed and execution effect of confidential documents.

[0061] In this embodiment, the core functions of the urgent reminder system include urgent detection, reminder triggering and information display. Generally, the system dynamically identifies urgent items through semantic analysis and status monitoring of the files to be processed, and transmits urgent information to users through voice prompts, alarms or highlighting. In some embodiments, this module further combines the circulation rules and node status of official documents to provide users with targeted processing suggestions.

[0062] Specifically, the urgent detection function uses the large model capability of the data processing module to deeply analyze the file content and extract the keywords and context information related to "urgent" in the file. The detection logic relies on the pre-defined urgent keyword library K = {k1, k2, ..., k n} and the file semantic feature weight W, the urgency of the file is scored by the following formula: Among them, P urgent Score the urgency of the file, α i is the keyword weight, Sim(v i ,k i ) represents the word vector v in the file i With keyword k i In one possible implementation, the system dynamically adjusts the keyword weight α i , in order to adapt to the processing needs of different types of urgent matters. For example, for files with the keywords "urgent", "time-limited" and "immediate", the system will prioritize marking them as urgent and trigger subsequent reminder mechanisms.

[0063] In this embodiment, the reminder trigger function automatically generates a reminder signal when specific conditions are met by real-time monitoring of the file status and user operations. As an option, the system uses the collaborative ability of the control module and the data processing module, combines the file's flow path and the current processing node, and determines whether the file has a risk of being unprocessed or overdue. The reminder trigger condition can be expressed by the following formula: R = f(P urgent ,t f ,t d ) Among them, R is the reminder signal, P urgent Score the urgency of the file, t f is the current time, t d is the deadline for processing the file. Function f uses logical rules to comprehensively judge the file score and time information to determine whether to trigger a reminder. As a possible implementation method, the reminder trigger function supports not only single reminders, but also multiple rounds of reminders. For example, for urgent documents that are not processed in time, the system will repeatedly send reminder signals at preset time intervals until the user confirms or completes the processing.

[0064] As a possible implementation method, the reminder trigger function supports not only a single reminder, but also multiple rounds of reminders. For example, for urgent items that are not processed in time, the system will repeatedly send reminder signals at preset time intervals until the user confirms or completes the processing. In this embodiment, the display function of reminder information supports multiple forms, including voice broadcast, interface highlight display, and alarm triggering. In general, the system presents urgent information to the user intuitively through a feedback module, and attaches key information of the file (such as title, sending unit, and deadline). In a possible implementation method, the system combines the touch screen of the self-service platform to display the details of the urgent item in a graphical interface, and marks the emergency status of the file with colors and icons. For example, the system highlights urgent files in red and adds an alarm symbol to increase the user's attention to the urgent item. The trigger logic of the alarm sound is as follows: A=g(P urgent ,Δt) Among them, A is the alarm signal, P urgent Score the urgency of the file, Δt = t f -t d It is the difference between the current time and the deadline. When Δt is less than the preset threshold, the system automatically triggers an alarm.

[0065] In some embodiments, the urgent reminder system also cooperates with the voice interaction module to inform the user of the urgent processing requirements through voice prompts. For example, when the user logs into the system, the voice interaction module will actively prompt "You have two urgent documents to be processed, the titles are..." This function not only improves the user experience, but also reduces the risk of delays caused by manual omissions.

[0066] In this embodiment, the synergy between the urgent reminder system and other modules ensures the accuracy and timeliness of the reminder information. By calling the status monitoring capability of the data processing module, the system can update the processing progress of the file in real time and adjust the reminder strategy; at the same time, through the control module to coordinate the voice interaction module and the self-service platform, the system can send reminders to users in multiple terminals and multiple scenarios.

[0067] In summary, the urgent reminder system has built an efficient and intelligent reminder mechanism through functional modules such as urgent detection, reminder triggering and information display. Its semantic parsing ability and priority discrimination ability based on the big model not only improve the timeliness of urgent processing, but also provide users with intuitive and reliable operation support, fully meeting the needs of urgent affairs processing in confidential document management.

[0068] Specific implementation method of the three-member management platform The three-member management platform in the intelligent voice interactive document exchange system based on the large model of the present invention is the core module to ensure the security, standardization and compliance of the system. The platform includes three parts: system administrator module, audit administrator module and security administrator module, which are respectively responsible for the allocation and management of system basic data and user permissions, the recording and analysis of system operation logs, and the security review and allocation verification of permission resources. The three-member management platform forms a full-process control of user operations and system operations through the collaborative work with the control module, data processing module and feedback module, ensuring the reliability and confidentiality of the system, while meeting the high security requirements for permission management in confidential document management.

[0069] In this embodiment, the functional division of the three-member management platform is based on the principle of separation of duties, which clarifies the responsibility boundaries between system administrators, audit administrators and security administrators, thereby reducing the security risks caused by concentrated authority. Under normal circumstances, the three-member management platform relies on the system authority model to restrict user operations, and combines the intelligent analysis capabilities of the large model to automatically detect and warn of abnormal operations, further improving the security and stability of the system.

[0070] In this embodiment, the main functions of the system administrator module include initialization of basic data, allocation of user rights and configuration management of system resources. Specifically, the system administrator module dynamically adjusts the user right information by calling the control module to ensure that the scope of each user's rights is consistent with his or her responsibilities. In some embodiments, the module uses a large model to intelligently analyze the matching relationship between user roles and system functions to generate an optimal rights allocation plan. For example, for a newly created user account U, the system administrator module calculates the appropriate scope of rights by the following formula: Among them, P(U) represents the permission set assigned to user U, is the permission set in the system, R(U) is the user's role information, and F(p) is the function set corresponding to permission p. By calculating the similarity between R(U) and F(p), the system can ensure the rationality of permission allocation.

[0071] As an option, the system administrator module also supports real-time review and optimization of permission allocation. For example, when a user's responsibilities change, the system will readjust the permission scope according to the new role requirements to ensure that the permission configuration is always consistent with actual needs.

[0072] In this embodiment, the function of the audit administrator module is mainly reflected in the recording and analysis of operation logs. Generally, the module performs in-depth analysis of the system operation logs by calling the data processing module, automatically detects and marks possible violations. In some embodiments, the audit administrator module uses the behavioral analysis capabilities of large models to identify abnormal patterns in operation logs and generate detailed audit reports. For example, when a user attempts to unauthorized access multiple times in a short period of time, the system will automatically record the relevant operations and trigger an audit alert.

[0073] Specifically, the audit administrator module uses the following formula to score the anomaly of the operation log: Among them, S anomaly is the abnormality score, L i represents the i-th log record, f(L i ) is the log anomaly detection function, w i is a weight factor, which indicates the influence of different log items on the anomaly score. anomaly When the preset threshold is exceeded, the system will mark the relevant log as high risk and send an alert to the audit administrator. In one possible implementation, the audit administrator module also supports the visual display function of the log. The system presents the analysis results of the log to the user in the form of a chart or table by calling the feedback module. For example, the system can generate a time-ordered abnormal operation frequency graph to help administrators quickly locate problematic operations. In this embodiment, the functions of the security administrator module include security audits and allocation verifications of permission resources to ensure that the allocation of user permissions is completely consistent with the system security policy. In general, the module performs a secondary review of the system administrator's permission configuration by calling the control module to prevent security vulnerabilities caused by permission abuse. In some embodiments, the security administrator module uses a large model to verify and analyze the permission allocation plan to generate a permission security score S. security , to evaluate the compliance of the allocation scheme. The calculation formula of the safety score is as follows: Among them, S security Score the security of the permission, p i represents the i-th permission, g(p i ,S) is the permission p i Compliance score with system security policy S, β i is the weight factor.

[0074] As an option, the security administrator module also supports retrospective checking of system permission configurations. For example, when an anomaly in the permission configuration is detected, the system automatically generates permission adjustment suggestions and sends a review request to the security administrator.

[0075] In one possible implementation, the three-member management platform works in conjunction with the voice interaction module to quickly execute management tasks through voice commands. For example, the audit administrator can use voice input to "query last week's high-risk operation logs", and the system will automatically retrieve relevant logs and generate analysis results. At the same time, the feedback module will return the results to the user in voice or graphical form, thereby improving operational efficiency.

[0076] Through the above functions, the three-member management platform realizes the full-process security management and control of system operations. Its intelligent analysis capabilities based on large models not only improve the efficiency of authority allocation and auditing, but also significantly enhance the security and stability of the system through anomaly detection and risk warning functions, providing a solid guarantee for the reliable operation of the confidential document exchange system.

[0077] Please refer to the attached Figure 2 The embodiment of the present invention also provides a human-computer dialogue control method for an intelligent voice interactive document exchange system based on a large model, comprising the following steps: The user inputs commands via voice; The system uses the voice interaction module to analyze the voice content and generate operation instructions based on the big model; The control module schedules the function modules in the system to execute the instructions; The feedback module outputs the operation results to the user in the form of voice or text.

[0078] in: The analysis step includes: using a large model to deeply analyze the semantics of the user's voice content, and generating accurate operation instructions in combination with the user's historical instructions and context information; The feedback step includes: Generate text content for the execution results and generate voice feedback through the speech synthesis module; The execution status is displayed on the touch screen of the self-service platform in a graphical interface for users to review.

[0079] The method of this embodiment is executed based on the above-mentioned system embodiment, and its principles and technical effects are similar, so they will not be repeated here.

[0080] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent voice interactive document exchange system based on a large model, characterized by: include: Confidential document management subsystem, used for document sending and receiving business management; Voice interaction module, used to receive and analyze user voice commands; The control module is used to schedule the sub-modules in the system to complete the operation instructions; The data processing module analyzes and processes the content, status and flow of official documents and letters based on the big model; The feedback module is used to feed back the execution results to the user in the form of voice or text.

2. The intelligent voice interactive document exchange system based on a large model according to claim 1 is characterized in that: The confidential document management subsystem automatically classifies and semantically analyzes document content based on a large model, and realizes intelligent management through keyword extraction and context analysis of text content. The large model is a pre-trained language model built based on deep learning technology. Through training on massive corpus, it has powerful semantic understanding and context analysis capabilities, and can perform accurate classification and semantic analysis on document content.

3. The intelligent voice interactive document exchange system based on a large model according to claim 1 is characterized in that: The voice interaction module uses a pre-trained language model to perform intent recognition and context analysis on the user's voice, generates operation instructions corresponding to the voice content, and schedules the corresponding functional modules to execute through the control module.

4. The intelligent voice interactive document exchange system based on a large model according to claim 1 is characterized in that: The feedback module comprises: A feedback generation unit, used for converting the execution result into a voice or graphical interface; The prompt unit is used to attract the user's attention through alarms or highlighted prompts when urgent items are being processed.

5. The intelligent voice interactive document exchange system based on a large model according to claim 1 is characterized in that: The data processing module extracts document semantic information through a pre-trained language model, and automatically classifies, sorts, and tracks the status of official documents in combination with the structural features of the official documents.

6. The intelligent voice interactive document exchange system based on a large model according to claim 1 is characterized in that: The system further comprises: Self-service platform system, installed in the exchange room, with functions including mail registration and barcode printing for confidential mail receiving and dispatching personnel; The system department query system is used by the confidential mail clerks of the networked departments to conduct sending and receiving inquiries and statistical management of the department's official documents and letters.

7. The intelligent voice interactive document exchange system based on a large model according to claim 1 is characterized in that: The system further comprises: Urgent reminder system, including: Urgent item detection unit, used to monitor the status of urgent items waiting to be picked up in the exchange box; A reminder trigger unit, for generating a reminder signal after detecting an urgent item; A reminder display unit, used to display reminder information on a user terminal or a self-service device in an exchange room; The three-member management platform includes: System administrator module, used to allocate and manage system basic data and user resource permissions; Audit administrator module, used to record and analyze system operation logs and check for violations; The security administrator module is used to review the permission resources allocated by the system administrator and ensure the security of user permissions.

8. A human-computer dialogue control method for an intelligent voice interactive document exchange system based on a large model, applied to the intelligent voice interactive document exchange system based on a large model as claimed in any one of claims 1 to 7, characterized in that: The following steps are involved: The user inputs commands via voice; The system uses the voice interaction module to analyze the voice content and generate operation instructions based on the big model; The control module schedules the function modules in the system to execute the instructions; The feedback module outputs the operation results to the user in the form of voice or text.

9. The human-computer dialogue control method of the intelligent voice interactive document exchange system based on a large model according to claim 8 is characterized in that: The analysis step includes: using a large model to deeply analyze the semantics of the user's voice content, and generating accurate operation instructions in combination with the user's historical instructions and context information.

10. The human-computer dialogue control method of the intelligent voice interactive document exchange system based on a large model according to claim 8 is characterized in that: The feedback step includes: Generate text content for the execution results and generate voice feedback through the speech synthesis module; The execution status is displayed on the touch screen of the self-service platform in a graphical interface for users to review.

Images