Complex text interpretation system and method based on large language model proxy

Abstract

The invention discloses a complex text interpretation system and method based on a large language model agent, and relates to the technical field of natural language understanding, the system comprises a file analysis module, an interpretation module and a risk control module; the file analysis module is used for analyzing the received complex file and identifying text content and non-text content in the complex file; non-text content is filtered and labeled, so that it is guaranteed that output complex text content is complete and clear; the interpretation module is connected with the file analysis module, a multi-agent interpretation architecture is adopted to generate a preliminary interpretation result, and the multi-agent interpretation architecture comprises a plurality of large language model agents and can interpret the complex text content in a multi-view collaborative mode; and the risk control module is connected with the interpretation module, and is used for verifying and optimizing potential deviation, ambiguity and logic inconsistency in the preliminary interpretation result to generate final interpretation output.

Description

Technical Field

[0001] The present invention relates to the technical field of natural language understanding, and in particular to a complex text interpretation system and method based on a large language model agent. Background Art

[0002] With the development of big data and artificial intelligence technology, intelligent interpretation systems based on natural language processing have been widely used in the field of text analysis. Large Language Models (LLMs), as the core technology in the field of natural language processing, have gradually demonstrated strong adaptability and accuracy in a variety of complex tasks with their powerful semantic understanding and generation capabilities. Through large-scale text training, large language models can achieve high levels of understanding and generation in tasks such as text summarization, language translation, and question answering, and are therefore widely used in various text-intensive tasks.

[0003] The intelligent agent system based on the large language model is an innovative application in recent years. The intelligent agent system refers to a system in which multiple autonomous large language model agents work together to complete tasks. Through the collaborative work between different agents, the intelligent agent system can effectively integrate the understanding and analysis capabilities of each agent, and apply it to interpret complex and professional text information to form systematic and structured output results. The collaborative interpretation between agents enables the system to perform more comprehensive and accurate interpretations under a multi-perspective analysis and multi-level progressive architecture.

[0004] However, the traditional way of interpreting complex texts still relies mainly on manual analysis. This method not only consumes a lot of human resources, but also the accuracy and consistency of the interpretation cannot be guaranteed. Complex texts usually have rigorous terminology and complex expressions. The personal background, professional level and understanding angle of the interpreter often affect the understanding of the text, resulting in differences and subjectivity in the interpretation results. Different interpreters may have different interpretations of the same text, which may lead to unnecessary disagreements or even disputes in the interpretation of texts involving multiple stakeholders. In addition, the interests involved in the text are often relatively broad, and it is impossible to ensure that the results of manual interpretation always meet the requirements of fairness and justice.

[0005] In the current technical context, although some complex text interpretation systems based on large language models have been explored and developed, these systems are mostly focused on understanding a single text and lack a multi-agent collaborative interpretation architecture. When faced with the multi-level content of complex documents, the existing single-agent model often has misunderstandings due to the lack of a structured interpretation method, especially when it comes to the interpretation of detailed content such as laws and contracts, the performance of the existing system is still limited.

[0006] Therefore, technicians in this field are committed to developing a new complex text interpretation system and method to solve the above-mentioned defects in the prior art. Summary of the invention

[0007] In view of the above-mentioned defects of the prior art, the technical problem to be solved by the present invention is how to solve the problems of low efficiency and large interpretation differences in complex text interpretation.

[0008] To achieve the above-mentioned purpose, the present invention provides a complex text interpretation system based on a large language model agent, including a file parsing module, an interpretation module and a risk control module; The file parsing module parses the received complex file to identify the text content and non-text content therein; for the non-text content, filtering and marking processing are performed to ensure that the output complex text content is complete and clear; The interpretation module is connected to the file parsing module and generates a preliminary interpretation result using a multi-agent interpretation architecture, wherein the multi-agent interpretation architecture includes a plurality of large language model agents, which can interpret the complex text content in a collaborative manner from multiple perspectives; The risk control module is connected to the interpretation module to verify and optimize potential deviations, ambiguities and logical inconsistencies in the preliminary interpretation results to generate a final interpretation output.

[0009] Furthermore, the file parsing module also formats and segments the complex text content to improve the orderliness and readability of the data.

[0010] Furthermore, the interpretation module includes two interpretation architectures: a peer system and a hierarchical system.

[0011] Furthermore, in the same-level system, the multiple large language model agents are deployed in parallel, and each large language model agent has a different role definition and is responsible for interpretation tasks from different perspectives.

[0012] Furthermore, in the hierarchical system, the multiple large language model agents are divided into low-level agents and high-level agents, which transmit and interpret the content layer by layer in a progressive manner. The low-level agents are weakly related role definitions, responsible for the preliminary parsing and summarization of basic information, and the high-level agents are strongly related role definitions, which perform deeper analysis and judgment based on the basic information.

[0013] The present invention also provides a complex text interpretation method based on a large language model agent, the method comprising the following steps: Step 1: parse the received complex file to identify the text content and non-text content; filter and annotate the non-text content to ensure that the output complex text content is complete and clear; Step 2: Generate preliminary interpretation results using a multi-agent interpretation framework, wherein the multi-agent interpretation framework includes multiple large language model agents that can interpret the complex text content in a collaborative manner from multiple perspectives; Step 3: Verify and optimize potential deviations, ambiguities and logical inconsistencies in the preliminary interpretation results to generate a final interpretation output.

[0014] Furthermore, the multi-agent interpretation architecture in step 2 is divided into two interpretation architectures: a peer system and a hierarchical system.

[0015] Furthermore, in the same-level system, the multiple large language model agents are deployed in parallel, and each large language model agent has a different role definition and is responsible for interpretation tasks from different perspectives.

[0016] Furthermore, in the hierarchical system, the multiple large language model agents are divided into low-level agents and high-level agents, which transmit and interpret the content layer by layer in a progressive manner. The low-level agents are weakly related role definitions, responsible for the preliminary parsing and summarization of basic information, and the high-level agents are strongly related role definitions, which perform deeper analysis and judgment based on the basic information.

[0017] Furthermore, the step 3 includes the following sub-steps: Step 3.1: Perform consistency check by comparing the interpretation results of each agent, identifying the interpretation differences, and ensuring that the output results are highly consistent based on the analysis of multiple agents; Step 3.2: Perform bias correction, mark and adjust sensitive content in the interpretation, and ensure that the output results are objective and fair; Step 3.3: Optimize the results, generate optimized interpretation results and output the final interpretation output that meets semantic consistency and logical rigor.

[0018] The complex text interpretation system and method based on a large language model agent provided by the present invention has at least the following technical effects: 1. The technical solution provided by the present invention effectively makes up for the shortcomings of manual interpretation methods. It adopts a multi-agent collaborative interpretation architecture and can analyze the content of complex texts at multiple levels and from multiple perspectives through the design of peer systems and hierarchical systems. Through mutual discussions between peer agents and the progressive level of hierarchical agents, the interpretation results generated by the system are not only more comprehensive, but also highly consistent and logical. This design greatly reduces the errors caused by personal interpretation bias and ensures the objectivity and fairness of the interpretation process.

[0019] 2. The technical solution provided by the present invention also performs strict verification and optimization after generating the interpretation results, making the output results more accurate and reliable. Through steps such as consistency detection and deviation correction, potential misunderstandings or ambiguities can be effectively identified and corrected to ensure the logical clarity and accuracy of the output content. In addition, the technical solution provided by the present invention also provides a special processing mechanism for sensitive information in complex texts, ensuring the reliability of the interpretation results in text interpretation tasks with high precision requirements.

[0020] In summary, the technical solution provided by the present invention has significant advantages in the accuracy, consistency and efficiency of complex text interpretation. Compared with traditional manual interpretation methods, it significantly improves the quality and efficiency of text interpretation, and provides solid technical support for the systematic interpretation and intelligent application of complex texts.

[0021] The concept, specific structure and technical effects of the present invention will be further described below in conjunction with the accompanying drawings to fully understand the purpose, characteristics and effects of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 is a flow chart of a complex text interpretation method according to a preferred embodiment of the present invention; Figure 2 It is a schematic diagram of the interpretation architecture of a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0023] The following describes several preferred embodiments of the present invention with reference to the drawings in the specification, so that the technical content is clearer and easier to understand. The present invention can be embodied in many different forms of embodiments, and the protection scope of the present invention is not limited to the embodiments mentioned in the text.

[0024] Traditional interpretation of complex texts mainly relies on manual analysis, which is often inefficient and easily affected by the interpreter's personal subjective factors when facing complex texts, resulting in inconsistencies in the interpretation results. Manual interpretation is easily affected by the interpreter's knowledge background, interest position and reading habits, resulting in large differences in the understanding of the same complex text by different interpreters, which is particularly evident in the interpretation of texts involving multiple interests. In addition, due to the high professionalism and legality of complex texts, manual interpretation may miss key information and is difficult to achieve efficient and accurate requirements. In response to the problems of low efficiency and large interpretation differences in the interpretation of complex texts, an embodiment of the present invention proposes a text interpretation system based on a large language model agent. The system combines the semantic understanding and generation capabilities of the large language model and designs a multi-agent interpretation architecture to achieve systematic, automated and multi-perspective interpretation of complex texts, thereby effectively improving the consistency and accuracy of the interpretation.

[0025] Example 1 like Figure 1 As shown, a complex text interpretation system based on a large language model agent is provided in an embodiment of the present invention, including a file parsing module, an interpretation module and a risk control module, which can adapt to texts of different types and complexities and ensure high-quality output of interpretation results.

[0026] The file parsing module parses the received complex files and identifies the text and non-text contents therein; it filters and annotates non-text contents such as charts and annotations to ensure that the output complex text contents are complete and clear; The interpretation module is connected to the file parsing module and uses a multi-agent interpretation architecture to generate preliminary interpretation results. The multi-agent interpretation architecture includes multiple large language model agents, which can interpret complex text content in a collaborative way from multiple perspectives. The risk control module is connected to the interpretation module to verify and optimize the potential deviations, ambiguities and logical inconsistencies in the preliminary interpretation results and generate the final interpretation output.

[0027] In particular, the file parsing module receives the file in the initial format and pre-processes the file content so that the subsequent modules can interpret it effectively. The text content is identified, mainly some structured information such as titles, paragraphs, clauses, etc., and orderly text data is generated.

[0028] In particular, the file parsing module also formats and segments complex text content to improve the orderliness and readability of the data and provide high-quality text input for the interpretation module.

[0029] Example 2 like Figure 2As shown, based on Example 1, the interpretation module includes two interpretation architectures: a peer system and a hierarchical system.

[0030] In the same-level system, multiple large language model agents are deployed in parallel, and the agents collaborate in parallel. Each large language model agent has a different role definition and is responsible for interpretation tasks from different perspectives. They reach consensus interpretation through mutual discussion, which is suitable for processing more complex text content that requires multi-dimensional analysis. For example, one agent is responsible for analyzing the terms from a citizen's perspective, while another agent focuses on the impact on related industries. During the interpretation process, the agents discuss and exchange information with each other to reach a consensus on the content, thereby generating a comprehensive interpretation result with high consistency.

[0031] In the hierarchical system, multiple large language model agents are divided into low-level agents and high-level agents, which pass on the interpretation content layer by layer in a progressive manner. Among them, the low-level agents are weakly related role definitions, responsible for the preliminary analysis and summary of basic information, and the high-level agents are strongly related role definitions, which conduct deeper analysis and judgment based on the basic information. This architecture is particularly suitable for processing multi-level and logically complex files, which helps to improve the orderliness and logic of the interpretation process. The hierarchical system structure allows agents to progress layer by layer, gradually deepening the understanding and analysis of the text, and ensuring the clear transmission and logical progression of information flow. Such a multi-agent collaborative architecture not only solves the subjectivity and inconsistency problems that may exist in traditional interpretation methods, but also significantly improves the efficiency of interpretation. It is suitable for complex text interpretation tasks that require rapid understanding and feedback.

[0032] Example 3 On the basis of Example 1 or 2, in order to ensure the accuracy and applicability of the interpretation results, a risk control module is designed in the embodiment of the present invention. The risk control module performs multiple evaluations on the interpretation results after they are generated, identifies possible deviations and ambiguities, and makes necessary adjustments and optimizations based on the background and context of the text content to ensure that the final output meets the actual application requirements. By introducing the contextual understanding and multi-level collaborative capabilities of the large language model agent, the embodiment of the present invention effectively improves the accuracy and consistency of text interpretation, and provides intelligent and efficient support for professionals and the public in related fields in the understanding and application of complex texts.

[0033] Specifically, after generating the preliminary interpretation results, the risk control module conducts final verification and optimization of the interpretation content. The risk control module uses the context analysis capabilities of the large language model to review potential biases, ambiguities, and logical inconsistencies in the interpretation results. Specifically, it includes: 1. Consistency check: By comparing the interpretation results of each agent, potential interpretation differences are identified to ensure that the output results are highly consistent based on the analysis of multiple agents.

[0034] 2. Bias correction: For sensitive content in interpretation (such as explanation and influence of stakeholders), the risk control module marks and adjusts possible biases to ensure that the output results are objective and fair.

[0035] 3. Result optimization: After verification and adjustment, the system generates optimized interpretation results and outputs a final interpretation report that meets semantic consistency and logical rigor.

[0036] Through the collaborative work of various modules, the embodiment of the present invention provides a systematic and intelligent document interpretation system, which is particularly suitable for multi-level analysis and multi-angle interpretation of complex documents. The system can not only ensure the accuracy and consistency of the interpretation content, but also significantly improve the interpretation efficiency, providing efficient document understanding support for enterprises and the public.

[0037] Example 4 The embodiment of the present invention provides a complex text interpretation method based on a large language model agent, comprising the following steps: Step 1: parse the received complex file to identify the text content and non-text content; filter and annotate the non-text content to ensure that the output complex text content is complete and clear; Step 2: Generate preliminary interpretation results using a multi-agent interpretation architecture, where the multi-agent interpretation architecture includes multiple large language model agents that can interpret complex text content in a collaborative manner from multiple perspectives; Step 3: Verify and optimize potential biases, ambiguities, and logical inconsistencies in the preliminary interpretation results to generate the final interpretation output.

[0038] Example 5 Based on Example 4, the multi-agent interpretation architecture in step 2 is divided into two interpretation architectures: a peer system and a hierarchical system.

[0039] In the same-level system, multiple large language model agents are deployed in parallel. Each large language model agent has different role definitions and is responsible for interpretation tasks from different perspectives.

[0040] In the hierarchical system, multiple large language model agents are divided into low-level agents and high-level agents, which pass on the interpretation content layer by layer in a progressive manner. Among them, the low-level agent is a weakly related role definition, responsible for the preliminary parsing and summarization of basic information, and the high-level agent is a strongly related role definition, which conducts deeper analysis and judgment based on the basic information.

[0041] Example 6 Based on Example 4 or 5, step 3 includes the following sub-steps: Step 3.1: Perform consistency check by comparing the interpretation results of each agent, identifying the interpretation differences, and ensuring that the output results are highly consistent based on the analysis of multiple agents; Step 3.2: Perform bias correction, mark and adjust sensitive content in the interpretation, and ensure that the output results are objective and fair; Step 3.3: Optimize the results, generate optimized interpretation results and output the final interpretation output that meets semantic consistency and logical rigor.

[0042] The preferred specific embodiments of the present invention are described in detail above. It should be understood that ordinary technicians in the field can make many modifications and changes based on the concept of the present invention without creative work. Therefore, all technical solutions that can be obtained by technicians in the technical field based on the concept of the present invention through logical analysis, reasoning or limited experiments on the basis of the prior art should be within the scope of protection determined by the claims.

Claims

1. A complex text interpretation system based on a large language model agent, characterized in that: Includes file parsing module, interpretation module and risk control module; The file parsing module parses the received complex file to identify the text content and non-text content therein; for the non-text content, filtering and marking processing are performed to ensure that the output complex text content is complete and clear; The interpretation module is connected to the file parsing module and generates a preliminary interpretation result using a multi-agent interpretation architecture, wherein the multi-agent interpretation architecture includes a plurality of large language model agents, which can interpret the complex text content in a collaborative manner from multiple perspectives; The risk control module is connected to the interpretation module to verify and optimize potential deviations, ambiguities and logical inconsistencies in the preliminary interpretation results to generate a final interpretation output.

2. The complex text interpretation system based on a large language model agent as claimed in claim 1, characterized in that: The file parsing module also formats and segments the complex text content to improve the orderliness and readability of the data.

3. The complex text interpretation system based on a large language model agent as claimed in claim 1, characterized in that: The interpretation module includes two interpretation architectures: a peer system and a hierarchical system.

4. The complex text interpretation system based on a large language model agent as claimed in claim 3, characterized in that: In the same-level system, the multiple large language model agents are deployed in parallel, each of which has a different role definition and is responsible for interpretation tasks from different perspectives.

5. The complex text interpretation system based on a large language model agent as claimed in claim 3, characterized in that: In the hierarchical system, the multiple large language model agents are divided into low-level agents and high-level agents, which transmit and interpret the content layer by layer in a progressive manner. Among them, the low-level agents are weakly related role definitions, responsible for the preliminary parsing and summarization of basic information, and the high-level agents are strongly related role definitions, which perform deeper analysis and judgment based on the basic information.

6. A complex text interpretation method based on a large language model agent, characterized in that: The method comprises the following steps: Step 1: parse the received complex file to identify the text content and non-text content; filter and annotate the non-text content to ensure that the output complex text content is complete and clear; Step 2: Generate preliminary interpretation results using a multi-agent interpretation framework, wherein the multi-agent interpretation framework includes multiple large language model agents that can interpret the complex text content in a collaborative manner from multiple perspectives; Step 3: Verify and optimize potential deviations, ambiguities and logical inconsistencies in the preliminary interpretation results to generate a final interpretation output.

7. The complex text interpretation method based on a large language model agent as claimed in claim 6, characterized in that: The multi-agent interpretation architecture in step 2 is divided into two interpretation architectures: a peer system and a hierarchical system.

8. The complex text interpretation method based on a large language model agent as claimed in claim 7, characterized in that: In the same-level system, the multiple large language model agents are deployed in parallel, each of which has a different role definition and is responsible for interpretation tasks from different perspectives.

9. The complex text interpretation method based on a large language model agent according to claim 7, characterized in that: In the hierarchical system, the multiple large language model agents are divided into low-level agents and high-level agents, which transmit and interpret the content layer by layer in a progressive manner. Among them, the low-level agents are weakly related role definitions, responsible for the preliminary parsing and summarization of basic information, and the high-level agents are strongly related role definitions, which perform deeper analysis and judgment based on the basic information.

10. The complex text interpretation method based on a large language model agent according to claim 6, characterized in that: The step 3 includes the following sub-steps: Step 3.1: Perform consistency check by comparing the interpretation results of each agent, identifying the interpretation differences, and ensuring that the output results are highly consistent based on the analysis of multiple agents; Step 3.2: Perform bias correction, mark and adjust sensitive content in the interpretation, and ensure that the output results are objective and fair; Step 3.3: Optimize the results, generate optimized interpretation results and output the final interpretation output that meets semantic consistency and logical rigor.

Images