Intelligent reasoning multiplexing method and system based on business analysis template library

Abstract

The invention provides an intelligent reasoning multiplexing method and system based on a business analysis template library, and solves the technical problems of reasoning redundancy, low efficiency and incapability of multiplexing knowledge when tasks with high repeatability and similar logic structures are processed in the prior art. The method comprises the steps of obtaining a user request for analysis to obtain a request intention; judging whether the request intention has a matched service analysis template or not, and if yes, loading the service analysis template; otherwise, splitting the request intention into a plurality of sub-tasks, splitting the sub-tasks into a plurality of query questions, circularly obtaining intermediate conclusions corresponding to all the query questions until all the sub-tasks are completed, summarizing all the intermediate conclusions, and outputting a reasoning report; judging whether the request intention is successfully completed for the first time or not, and if yes, packaging the reasoning path to obtain a new service analysis template; otherwise, finishing the reasoning process. The method can be widely applied to the technical field of business analysis.

Description

Technical Field

[0001] This application belongs to the field of business analysis technology, and more specifically, it relates to an intelligent reasoning reuse method and system based on a business analysis template library. Background Technology

[0002] In professional fields such as financial investment research, insurance actuarial science, and government consulting, users often need to handle a large number of complex tasks with similar characteristics. For example, when conducting family protection gap analysis or providing retirement planning advice for different clients, the underlying logic is highly consistent: first, collect basic information; then, analyze existing protection; next, compare it with target needs; and finally, propose supplementary solutions.

[0003] However, existing AI systems often handle such tasks using a step-by-step reasoning approach. Each request requires starting from scratch to understand, break down, call tools, and analyze results, repeating the entire thought process even if the problem structure is exactly the same. This leads to long reasoning times, low resource utilization, poor consistency in responses to similar problems, inability to accumulate experience for high-frequency tasks, and difficulty in knowledge accumulation, hindering the formation of organizational-level intelligent assets. Although some research has attempted to introduce memory mechanisms, such as long-term memory and vector databases, these mainly focus on fact storage, such as customer age and insurance coverage. They lack the ability to systematically preserve and reuse reasoning paths and analytical structures. Therefore, a technical solution capable of capturing and reusing complete reasoning patterns is urgently needed to address these problems. Summary of the Invention

[0004] The purpose of this application is to provide an intelligent reasoning reuse method and system based on a business analysis template library, so as to solve the technical problems of redundant reasoning, low efficiency and inability to reuse knowledge when the prior art is dealing with highly repetitive tasks with similar logical structures.

[0005] To achieve the above objectives, a first aspect of this application provides an intelligent reasoning reuse method based on a business analysis template library, comprising the following steps: The user request is obtained and parsed to determine the request intent; Determine if there is a matching business analysis template for the request intent. If so, load the business analysis template. Otherwise, break down the request intent into several sub-tasks, decompose the sub-tasks into several query questions, and iterate to obtain the intermediate conclusions corresponding to all query questions until all sub-tasks are completed. Summarize all intermediate conclusions and output a reasoning report. Determine if this is the first successful completion of the request intent. If so, encapsulate the reasoning path and obtain a new business analysis template; otherwise, complete the reasoning process.

[0006] Preferably, the process of obtaining intermediate conclusions corresponding to the query question includes: confirming the execution information of the query question, obtaining the original data, performing completion reasoning, text analysis and logical judgment, and then combining it with a pattern database or expert knowledge to generate intermediate conclusions corresponding to the query question.

[0007] Preferably, the execution information includes the query parameters for the query question, server_name, and tools. The query parameter is used to obtain the main parameters of the query question; server_name is used to obtain the data source address for the query question; The tools toolkit is used to obtain specific functional modules for executing queries and processing data.

[0008] Preferably, obtaining raw data refers to obtaining the main parameters of the query question, the data source address, and the specific data to be queried through execution information.

[0009] Preferably, the process of performing completion reasoning, text analysis, and logical judgment includes: filling the original data into preset prompt words; performing completion reasoning on the missing information in the original data according to the task instructions in the original data and preset prompt words to obtain complete information; performing text analysis on the complete information, extracting keywords, and performing logical judgment on the complete information based on the keywords to generate analysis results.

[0010] Preferably, the process of determining whether there is a matching business analysis template for the request intent includes: performing template retrieval of the request intent in the template library based on a semantic similarity algorithm, and determining whether there is a business analysis template that matches the request intent through vector search.

[0011] Preferably, the process of obtaining the request intent includes: performing semantic understanding and intent recognition on the user request to obtain the request intent; The process of semantic understanding and intent recognition includes: parsing the user request, extracting the target of the user request, decomposing the target into several sub-tasks, analyzing them to obtain the sub-task intents, integrating all sub-task intents, and clarifying the request intent.

[0012] Preferably, the business analysis template includes standardized analysis templates generated during the reasoning process. Each analysis template includes a template ID, trigger keywords, semantic features, analysis logic structure, tool list, output format, historical usage records, and confidence score.

[0013] A second aspect of this application provides an intelligent reasoning reuse system based on a business analysis template library, including: a planner, an executor, and an analyzer; The planner is used to obtain user requests, parse them to get the request intent, and determine whether there is a matching business analysis template for the request intent. If there is, the executor loads the business analysis template; otherwise, the request intent is broken down into several sub-tasks and input into the executor. The executor is used to break down subtasks into several query questions and send them to the analyzer; The analyzer is used to iterate through all the intermediate conclusions for all the queries, summarize all the intermediate conclusions, and output a reasoning report.

[0014] Preferably, the template generation module is used to determine whether it is the first successful completion of the request intent. If so, it triggers the template update module to encapsulate the reasoning path and obtain a new business analysis template; otherwise, it completes the reasoning process.

[0015] The beneficial effects of this application are as follows: This application provides an intelligent reasoning reuse method and system based on a business analysis template library. By parsing user requests into request intents and determining whether there is a matching business analysis template based on the request intents, the completed reasoning path corresponding to the request intents is solidified into a standardized business analysis template. This enables rapid response and reasoning reuse for subsequent user requests, allowing the business analysis template to directly call the corresponding reasoning path when processing similar user requests in the future, without having to re-execute the complete reasoning process, significantly improving response efficiency and result consistency. After successfully completing a certain type of complex task for the first time, the system can automatically seal the corresponding request reasoning path and save it to the business analysis template library. Subsequent user requests can directly match the corresponding reasoning path through the request intent, forming an intelligent reasoning closed loop of one-time construction and multiple reuse, reducing reasoning redundancy during computation and efficiently realizing reasoning reuse. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 A schematic diagram of the overall process of an intelligent reasoning reuse method based on a business analysis template library provided in an embodiment of this application; Figure 2 This is a schematic diagram of the overall process of an intelligent reasoning reuse system based on a business analysis template library, provided as an embodiment of this application. Detailed Implementation

[0018] To make the technical problems, technical solutions, and beneficial effects to be solved by this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and are not intended to limit the scope of this application.

[0019] Please see Figure 1 The first embodiment of this application provides an intelligent reasoning reuse method based on a business analysis template library, comprising: S1: Obtain the user request, parse it, and obtain the request intent.

[0020] Specifically, when a user inputs a natural language request, the planner receives the request and, based on preset prompts, uses a large language model to perform semantic understanding and intent recognition on the request to obtain the request intent.

[0021] In an optional embodiment, the user inputs a natural language request, "Help me analyze my family's security gap," which is then passed to the planner. The planner fills the received request, "Provide me with a report on my family's security gap," into a preset prompt template. First, it parses the request, "Provide me with a report on my family's security gap," extracting the user's core objective. In this embodiment, the core objective is "Provide a report on my family's security gap." The core objective is systematically broken down into a series of logically coherent subtasks, "analysis" and "family security gap." The intentions of the subtasks are analyzed, and all subtask intentions are integrated to clarify the request intention: "Analyze the family's security gap."

[0022] S2: Determine if there is a matching business analysis template for the request intent. If so, load the template.

[0023] Specifically, the planner performs template retrieval on the parsed request intent in the business analysis template library using a semantic similarity algorithm (such as a BERT-based text embedding similarity algorithm). It uses vector search to determine if a matching business analysis template exists. If a corresponding template exists and its confidence level meets the requirements, the executor directly loads the template and obtains a reasoning report for the request intent based on the reasoning process of the existing template. The business analysis templates include standardized analysis templates generated during historical reasoning. Each template includes a template ID, triggering keywords / semantic features, analysis logic structure, a list of required tools, output format specifications, historical usage records, and a confidence score.

[0024] In an optional embodiment, the planner performs template retrieval for the request intent "analyze family security gaps" using a semantic similarity algorithm, transforming the request intent "analyze family security gaps" into a high-dimensional semantic vector. Then, it performs vector similarity matching on the request intent "analyze family security gaps" in a pre-defined business analysis template library. Specifically, the planner calculates the cosine similarity between the request intent vector and the trigger semantic feature vectors of each template in the template library, selecting the template with the highest similarity. Once the similarity between the "Family Security Gap Special Analysis Template" and the request intent "analyze family security gaps" reaches a preset threshold, and the template's confidence score meets the system call criteria, the executor directly loads the template. Then, the executor, according to the template's built-in analysis logic structure, sequentially calls the necessary tools such as the family financial data collection tool and risk calculation tool, following the output format specifications stipulated in the template and the historical reasoning process experience associated with the template, ultimately generating a complete reasoning report for the request intent "analyze family security gaps".

[0025] S3: If there is no business analysis template that matches the request intent, break down the request intent into several sub-tasks, decompose the sub-tasks into several query questions, confirm the execution information of the query questions, call external services to obtain raw data, perform completion reasoning, text analysis and logical judgment, combine with the rule base or expert knowledge to generate intermediate conclusions, iterate to obtain the intermediate conclusions corresponding to all query questions, summarize and output the reasoning report.

[0026] Specifically, when the planner fails to find a business analysis template matching the request intent, it uses a large language model based on preset prompts to break down the request intent into subtasks, and then loads the executor. The executor first breaks down these high-level subtasks into specific query questions. For example, querying Zhang's three-year income or obtaining information about children's education expenses.

[0027] Then confirm the query parameters, server_name, tools, and other execution information for the specific query question. Finally, call an external service (such as MCP) to retrieve the raw data.

[0028] In an optional embodiment, the query "Query Zhang's income for three years" corresponds to the query parameters including the subject parameter "Zhang San", the time parameter "a specific year", and the range parameter "annual income". `server_name` refers to the data source address for the corresponding query task. In this embodiment, the query for Zhang's income for three years can be retrieved from the customer financial information management system. `tools` refers to the specific functional modules that execute queries and process data. In this embodiment, when querying Zhang's income for three years, the `tools` toolset consists of bank data interface calling tools and data aggregation and calculation tools, used to connect to the bank system, obtain Zhang San's bank statements, and extract the total annual income from the statements.

[0029] After acquiring the raw data, the analyzer fills in preset prompts with the raw data. Based on a large language model, it performs completion reasoning, text analysis, and logical judgment on the raw data to obtain the analysis results. Specifically, the analyzer parses the prompts after filling in the raw data, locates missing keywords, process steps, and other information in the raw data, and combines the task instructions in the prompts with reverse reasoning based on the existing raw data to fill in the information gaps in the prompts and obtain complete information. Text analysis is then performed on the completed information to extract keywords. Logical verification and judgment are then performed on the complete information based on the keywords to generate analysis results. The analysis results are combined with a rule base or expert knowledge to generate intermediate conclusions. This process is repeated until intermediate conclusions are obtained for all query questions. All intermediate conclusions are then summarized and a reasoning report is output.

[0030] S4: The system determines whether this is the first successful completion of the request intent. If so, it triggers the encapsulation of the reasoning path and obtains a new business analysis template; otherwise, it completes the reasoning process.

[0031] Finally, the system determines whether this is the first successful completion of this type of request intent. If so, it triggers the template generation module to modify the content planned in the planner, encapsulating the updated planning content (this request intent and the corresponding inference path) into a new business analysis template and storing it in the template library; otherwise, the process ends. When the system subsequently receives the same or similar request intent, it directly calls the stored business analysis template, achieving efficient reuse of the inference process and forming a closed-loop mechanism of initial inference, template accumulation, and subsequent calls, significantly improving system response speed and consistency.

[0032] Please see Figure 2 The second embodiment of this application provides an intelligent reasoning reuse system based on a business analysis template library, which includes: a planner, an executor, an analyzer, a template generation module, and a template update module.

[0033] The planner is used to acquire and parse user requests to obtain the request intent. It then determines if a matching business analysis template exists for the request intent. If so, it controls the executor to load the template and reason about the request intent based on the existing reasoning path, generating a reasoning report. Otherwise, it uses a large language model based on preset prompts to break down the request intent into subtasks, which are then loaded by the executor. The business analysis template library includes standardized analysis templates generated during historical reasoning. Each template includes a template ID, triggering keywords / semantic features, analysis logic structure (e.g., first query A, then B, and then compare), a list of required tools, output format specifications, historical usage records, and a confidence score.

[0034] The executor is used to load the corresponding business analysis template and execute the reasoning process when there is a corresponding business analysis template for the request intent; when there is no corresponding template, the subtask is decomposed into a query question, the execution information of the query question is confirmed (query parameters, server_name, tools, etc.), and the external service (MCP) is called to obtain the raw data.

[0035] The analyzer performs completion reasoning, text analysis, and logical judgment on the raw data, then combines this with a pattern database or expert knowledge to generate intermediate conclusions. It iteratively retrieves intermediate conclusions for all query questions until all subtasks are completed. After completing all subtasks, it summarizes all intermediate conclusions and outputs a reasoning report.

[0036] The template generation module is used to automatically abstract the key nodes, call order, and analysis logic of this reasoning process into a general template after the first completion of a certain type of request intent, and write it into the business analysis template library.

[0037] The template update module is used to optimize the content of the business analysis template library based on manual operation or automatic system operation, thereby improving matching accuracy and execution efficiency.

[0038] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0039] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for intelligent reasoning reuse based on a business analysis template library, characterized in that, Includes the following steps: The user request is obtained and parsed to determine the request intent; Determine if there is a matching business analysis template for the requested intent; if so, load the business analysis template. Otherwise, the request intent is broken down into several sub-tasks, the sub-tasks are further broken down into several query questions, and intermediate conclusions corresponding to all the query questions are obtained in a loop until all the sub-tasks are completed. All the intermediate conclusions are then summarized and a reasoning report is output. Determine whether this is the first successful completion of the requested intent. If so, encapsulate the reasoning path to obtain a new business analysis template. Otherwise, complete the reasoning process.

2. The intelligent reasoning reuse method based on a business analysis template library as described in claim 1, characterized in that, The process of obtaining the intermediate conclusion corresponding to the query question includes: confirming the execution information of the query question, obtaining the original data, performing completion reasoning, text analysis and logical judgment, and then combining it with the rule base or expert knowledge to generate the intermediate conclusion corresponding to the query question.

3. The intelligent reasoning reuse method based on a business analysis template library as described in claim 2, characterized in that, The execution information includes the query parameters for the query question, server_name, and tools toolset; The query parameter is used to obtain the main parameters of the query question; The server_name is used to obtain the data source address for the query question; The tools suite is used to obtain specific functional modules for performing queries and processing data.

4. The intelligent reasoning reuse method based on a business analysis template library as described in claim 3, characterized in that, Obtaining the original data refers to obtaining the main parameters of the query question, the data source address, and the specific data to be queried through the execution information.

5. The intelligent reasoning reuse method based on a business analysis template library as described in claim 3, characterized in that, The process of performing the completion reasoning, text analysis, and logical judgment includes: filling the original data into preset prompt words; performing completion reasoning on the missing information in the original data according to the task instructions in the original data and the preset prompt words to obtain complete information; performing text analysis on the complete information to extract keywords; and performing logical judgment on the complete information based on the keywords to generate analysis results.

6. The intelligent reasoning reuse method based on a business analysis template library as described in claim 1, characterized in that, The process of determining whether the request intent has a matching business analysis template includes: performing template retrieval on the request intent in the template library based on a semantic similarity algorithm, and determining whether there is a business analysis template that matches the request intent through vector search.

7. The intelligent reasoning reuse method based on a business analysis template library as described in claim 1, characterized in that, The process of obtaining the request intent includes: performing semantic understanding and intent recognition on the user request to obtain the request intent; The process of semantic understanding and intent recognition includes: parsing the user request, extracting the target of the user request, decomposing the target into several sub-tasks, analyzing them to obtain the sub-task intents, integrating all the sub-task intents, and clarifying the request intent.

8. The intelligent reasoning reuse method based on a business analysis template library as described in claim 1, characterized in that, The business analysis templates include standardized analysis templates generated during the reasoning process. Each analysis template includes a template ID, trigger keywords, semantic features, analysis logic structure, tool list, output format, historical usage records, and confidence score.

9. An intelligent reasoning reuse system based on a business analysis template library, applied to the intelligent reasoning reuse method based on a business analysis template library as described in any one of claims 1-8, characterized in that, include: Planner, executor, and analyzer; The planner is used to obtain user requests, parse them to get the request intent, and determine whether the request intent has a matching business analysis template. If so, the executor loads the business analysis template. Otherwise, the request intent is broken down into several sub-tasks and input into the executor; The executor is used to decompose the subtask into several query questions and send them to the analyzer; The analyzer is used to iteratively obtain all intermediate conclusions corresponding to the query questions, summarize all the intermediate conclusions, and output a reasoning report.

10. The intelligent reasoning reuse system based on a business analysis template library as described in claim 9, characterized in that, It also includes a template generation module and a template update module; The template generation module is used to determine whether the request intent is successfully completed for the first time. If so, the template update module is triggered to encapsulate the reasoning path and obtain a new business analysis template. Otherwise, complete the reasoning process.

Images