CN-122019638-A - Method and device for generating traceability report, electronic equipment and storage medium

CN122019638ACN 122019638 ACN122019638 ACN 122019638ACN-122019638-A

Abstract

The application discloses a method, a device, electronic equipment and a storage medium for generating a traceability report, which are used for determining a target task aiming at a research problem according to the research problem input by a user; the method comprises the steps of creating a traceability knowledge base with an association relation with a research problem, obtaining original text information of a step to be executed for executing a target task, determining metadata used for representing logical relation between the original text information and an execution result of the step to be executed, storing the original text information, the execution result and the metadata into the traceability knowledge base, and responding to a traceability request of a task report of a user for the target task, extracting the original text information, the execution result and the metadata from the traceability knowledge base to generate a traceability report. The method realizes the synchronous storage of the original information, the intermediate execution result and the logic association metadata thereof in the task generation process, constructs a complete mapping system from the actual manuscript to the final conclusion, and realizes the logic trace and evidence path restoration of the artificial intelligence generated content.

Inventors

LIU YONGPENG
Shao Ximeng
XU XIN
WANG SHIYI
LIU XIN
WAN GUOHUI
ZHANG RONGHUI
CUI YAO
JIANG SHAN

Assignees

泰康保险集团股份有限公司
泰康资产管理有限责任公司

Dates

Publication Date: 20260512
Application Date: 20251223

Claims (12)

1. A method for generating a traceability report, comprising: Determining a target task aiming at a research problem according to the research problem input by a user; Creating a traceability knowledge base with association relation with the research problems; acquiring original text information of a step to be executed for executing the target task; Determining metadata for characterizing a logical relationship between the original text information and an execution result of the step to be executed; storing the original text information, the execution result and the metadata to the traceability knowledge base; and responding to the received tracing request of the user for the task report of the target task, extracting the original text information, the execution result and the metadata from the tracing knowledge base to generate a tracing report.
2. The method of claim 1, wherein the traceability repository comprises an original material repository, a subtask data repository, and a metadata repository, and wherein storing the original text information, the execution results, and the metadata to the traceability repository comprises: storing the original text information into an original data knowledge base; storing the execution result to the subtask data knowledge base; and storing the metadata to the metadata knowledge base.
3. The method of claim 2, wherein the original knowledge base comprises a vector database and a structured search index, the step of storing the original text information to the original knowledge base comprising: Generating a time stamp of the original text information and acquiring a source path of the original text information; carrying out semantic slicing processing on the original text information to generate a plurality of original text blocks; Converting the original text block into a high-dimensional feature vector, and storing the high-dimensional feature vector into the vector database; Storing the original text block, the timestamp, and the source path to the structured search index.
4. A method according to claim 3, wherein the step of storing the execution results to the subtask data repository comprises: extracting an execution logic description and a preset output format corresponding to the step to be executed; Executing the step to be executed according to the preset output format to obtain an intermediate reasoning conclusion which is consistent with the execution logic description; Extracting key values and core fact fragments in the intermediate reasoning conclusion, and determining the key values and the core fact fragments as structural fact data; vectorization processing is carried out by adopting the intermediate reasoning conclusion, and a conclusion feature vector is generated; and storing the structured fact data and the conclusion feature vector together into a subtask data knowledge base.
5. The method of claim 4, wherein the step of storing the metadata to the metadata repository comprises: Determining a target tool used for executing the step to be executed when the step to be executed is executed; Recording a tool identifier of the target tool; Recording a call index between the step to be executed and an original text block corresponding to the step to be executed; Generating a tool execution time stamp for recording the tool identification calling action; and packaging the tool identifier, the call index and the tool execution time stamp into metadata used for representing the path mapping relation between the original text block and the execution result.
6. The method of claim 5, wherein the step of extracting the original text information, the execution result, and the metadata from the traceability knowledge base to generate a traceability report in response to receiving a traceability request of a user for a task report of the target task comprises: Responding to receiving a selection operation of a user for target data of a task report, and receiving a determination instruction of the user for the selection operation, and judging whether the selected target data is valid data or not; When the selected target data is judged to be not effective data, displaying prompt information for representing that the selected target data does not have an associated tracing result; And when the selected target data is judged to be effective data, extracting the original text information, the execution result and the metadata from the traceability knowledge base to generate a traceability report.
7. The method of claim 5, wherein the step of extracting the original text information, the execution result, and the metadata from the traceability knowledge base to generate a traceability report in response to receiving a traceability request of a user for a task report of the target task comprises: And responding to the received automatic execution tracing instruction sent by the user, representing that the whole task report is automatically traced after the task report is generated, and extracting the original text information, the execution result and the metadata from the tracing knowledge base to generate a tracing report.
8. The method according to claim 6 or 7, wherein the step of extracting the original text information, the execution result and the metadata from the traceability knowledge base to generate a traceability report comprises: matching corresponding logical relationship metadata from a metadata knowledge base by utilizing semantic blocks representing the target data or the whole task report; extracting a tracing text block from the original data knowledge base according to the logical relation metadata, and extracting a tracing execution result with a corresponding relation with the tracing text block from the subtask data knowledge base; and generating the traceability report based on the traceability text block and the traceability execution result.
9. The method of claim 8, further comprising, after the step of extracting the original text information, the execution result, and the metadata from the traceability knowledge base to generate a traceability report in response to receiving a traceability request of a user for a task report of the target task: responding to a verification request which is sent by a user and used for characterizing that when the generation of the traceability report is completed, the traceability report is subjected to verification operation, and calculating to obtain a credibility score by comparing semantic deviation of the traceability text block and the traceability execution result; When the reliability score is judged to be lower than a preset threshold value, determining an abnormal task node corresponding to the traceability text block according to the logic relation metadata; and re-executing the step to be executed, which has an association relation with the abnormal task node.
10. A traceability report generating device, comprising: The target task determining module is used for determining a target task aiming at a research problem according to the research problem input by a user; the traceability knowledge base creation module is used for creating a traceability knowledge base with an association relation with the research problem; the original text information acquisition module is used for acquiring original text information of a step to be executed for executing the target task; A metadata determining module for determining metadata for characterizing a logical relationship between the original text information and an execution result of the step to be executed; The traceability knowledge base storage module is used for storing the original text information, the execution result and the metadata into the traceability knowledge base; And the traceability report generation module is used for responding to a traceability request of a task report of the target task, which is received by a user, and extracting the original text information, the execution result and the metadata from the traceability knowledge base to generate a traceability report.
11. An electronic device comprising a processor, a memory and a program or instruction stored on the memory and executable on the processor, which program or instruction when executed by the processor implements the method of claims 1-7.
12. A readable storage medium, characterized in that it stores thereon a program or instructions, which when executed by a processor, implements the method according to claims 1-7.

Description

Method and device for generating traceability report, electronic equipment and storage medium Technical Field The present invention relates to the field of traceability report generation technologies, and in particular, to a traceability report generation method, a traceability report generation device, an electronic device, and a readable storage medium. Background With the wide application of the generated artificial intelligence technology in the professional fields of research, consultation and the like, an Agent can automatically disassemble tasks, retrieve data and generate a research report according to the problems of users. However, existing report generation techniques still face significant challenges in terms of accuracy and reliability. Firstly, when a traditional generation model generates a conclusion, a final text is often directly output, and a logic deduction process behind the conclusion cannot be recorded. It is difficult for a user to learn which piece of original material the conclusion was based on, which processing steps were performed, in the face of a specific value or point of view in the report, resulting in that the authority of the report is difficult to verify. Further, during complex task processing, there is a lack of explicit logical association records between the raw material, intermediate processing steps, and final conclusions. When a fact error or logic deviation (namely "model illusion") occurs in the report, the system fails to synchronously establish a mapping relationship from 'original material' to 'execution action' to 'output result' in the generation stage, so that a user or the system cannot accurately trace back the error point, and the logic consistency of the report content cannot be effectively verified. Finally, the tracing efficiency and granularity are insufficient. The existing simple retrieval technology can only provide related reference documents, cannot accurately achieve specific text blocks and the function of the specific text blocks in specific execution steps, so that a large amount of manual intervention is needed for secondary comparison in the traceability process, and real-time, automatic and deep verification requirements under strict scenes such as financial analysis are difficult to meet. Disclosure of Invention The embodiment of the invention provides a method, a device, electronic equipment and a readable storage medium for generating a traceability report, which are used for overcoming the problems or at least partially solving the problems. In order to solve the technical problems, the application is realized as follows: In a first aspect, an embodiment of the present application provides a method for generating a traceability report, including: Determining a target task aiming at a research problem according to the research problem input by a user; Creating a traceability knowledge base with association relation with the research problems; acquiring original text information of a step to be executed for executing the target task; Determining metadata for characterizing a logical relationship between the original text information and an execution result of the step to be executed; storing the original text information, the execution result and the metadata to the traceability knowledge base; and responding to the received tracing request of the user for the task report of the target task, extracting the original text information, the execution result and the metadata from the tracing knowledge base to generate a tracing report. Optionally, the traceability knowledge base includes an original knowledge base, a subtask data knowledge base and a metadata knowledge base, and the step of storing the original text information, the execution result and the metadata in the traceability knowledge base includes: storing the original text information into an original data knowledge base; storing the execution result to the subtask data knowledge base; and storing the metadata to the metadata knowledge base. Optionally, the original knowledge base includes a vector database and a structured search index, and the step of storing the original text information to the original knowledge base includes: Generating a time stamp of the original text information and acquiring a source path of the original text information; carrying out semantic slicing processing on the original text information to generate a plurality of original text blocks; Converting the original text block into a high-dimensional feature vector, and storing the high-dimensional feature vector into the vector database; Storing the original text block, the timestamp, and the source path to the structured search index. Optionally, the step of storing the execution result in the subtask data repository includes: extracting an execution logic description and a preset output format corresponding to the step to be executed; Executing the step to be executed according to the preset output