Search

CN-122021558-A - Summarization report generation method, device and equipment based on large language model

CN122021558ACN 122021558 ACN122021558 ACN 122021558ACN-122021558-A

Abstract

The application relates to the technical field of artificial intelligence and the technical field of project management, and discloses a summary report generation method, a device and equipment based on a large language model, wherein the method comprises the steps of selecting a plurality of preset features with similarity higher than a preset value as a plurality of preferred features, and taking knowledge texts associated with all the preferred features as a plurality of preferred knowledge texts; selecting a preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring the influence range of a reference item and the solution of the reference item from the target knowledge text, fusing the characteristics of the influence range of the reference item, the characteristics of the solution of the reference item and the splicing characteristics of the development data of the IPTV item, generating the fusion characteristics of the IPTV item, inputting the fusion characteristics into a large language model, and generating a summary report of the development data of the IPTV item through the large language model. The method and the device are beneficial to improving the generation efficiency of the summary report.

Inventors

  • CHEN LIPING

Assignees

  • 优地网络有限公司

Dates

Publication Date
20260512
Application Date
20260113

Claims (10)

  1. 1. A summary report method based on a large language model, which is applied to an electronic device, the summary report generating method comprising: Acquiring development data of an IPTV project, wherein the development data of the IPTV project comprises task information of the IPTV project, defect information of the IPTV project, progress information of the IPTV project and code submitting information of the IPTV project; Respectively extracting characteristics of task information of the IPTV project, defect information of the IPTV project, progress information of the IPTV project and code submission information of the IPTV project by adopting a text characteristic extraction model to respectively obtain the characteristics of the task information of the IPTV project, the defect information of the IPTV project, the progress information of the IPTV project and the code submission information of the IPTV project; Splicing the characteristics of the task information of the IPTV project, the characteristics of the defect information of the IPTV project, the characteristics of the progress information of the IPTV project and the characteristics of the code submitting information of the IPTV project to obtain the splicing characteristics of the development data of the IPTV project; Searching splicing features of the development data of the IPTV project in a knowledge base through a search enhancement generation model to obtain a search result, acquiring similarity between each preset feature of the knowledge base and the splicing features of the development data of the IPTV project in the search result, selecting a plurality of preset features with similarity higher than a preset value as a plurality of preferred features, and taking knowledge texts associated with all the preferred features as a plurality of preferred knowledge texts; Generating a comprehensive score of each preferred knowledge text according to the search frequency of each preferred knowledge text, the application frequency of each preferred knowledge text, the average feedback score of each preferred knowledge text, the front feedback rate of each preferred knowledge text and the comprehensive score model, selecting the preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring the influence range of a reference item and the solution of the reference item from the target knowledge text, fusing the characteristics of the influence range of the reference item, the characteristics of the solution of the reference item and the splicing characteristics of the development data of the IPTV item, generating the fusion characteristics of the IPTV item, inputting the fusion characteristics of the IPTV item into a large language model, and generating a summary report of the development data of the IPTV item through the large language model.
  2. 2. The summary report generating method according to claim 1, wherein the acquiring development data of an IPTV program, the summary report generating method comprising: and sending a data acquisition request to the project management platform, receiving a response message returned by the project management platform according to the data acquisition request, and acquiring the development data of the IPTV project from the response message.
  3. 3. The summary report generating method according to claim 1, wherein the retrieving, by retrieving the spliced features of the development data of the IPTV project in the knowledge base by using the enhanced generation model, obtains a retrieving result, and in the retrieving result, obtains a similarity between each preset feature of the knowledge base and the spliced feature of the development data of the IPTV project, selects a plurality of preset features having a similarity higher than a preset value as a plurality of preferred features, and uses knowledge texts associated with all the preferred features as a plurality of preferred knowledge texts, including: Loading a retrieval enhancement generation model through a model file, and transmitting the splicing characteristics of the development data of the IPTV project to the retrieval enhancement generation model; Searching the spliced features of the development data of the IPTV project in the knowledge base through a search enhancement generation model to obtain a search result, acquiring the similarity between each preset feature of the knowledge base and the spliced features of the development data of the IPTV project in the search result, selecting a plurality of preset features with the similarity higher than a preset value as a plurality of preferred features, and taking knowledge texts associated with all the preferred features as a plurality of preferred knowledge texts.
  4. 4. The summary report generating method according to claim 1, wherein the generating the comprehensive score of each preferred knowledge text based on the search frequency of each preferred knowledge text, the application frequency of each preferred knowledge text, the average feedback score of each preferred knowledge text, the front feedback rate of each preferred knowledge text, and the comprehensive score model, selecting a preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring the influence range of a reference item and the solution of the reference item from the target knowledge text, fusing the characteristics of the influence range of the reference item, the characteristics of the solution of the reference item, and the splice characteristics of the development data of the IPTV item, generating the fusion characteristics of the IPTV item, inputting the fusion characteristics of the IPTV item into a large language model, and generating the summary report of the development data of the IPTV item through the large language model, comprises: Acquiring the total number of times each preferred knowledge text is searched in preset time, selecting the total number of times each preferred knowledge text is searched in the preset time as the searching frequency of each preferred knowledge text, acquiring the total number of times each preferred knowledge text is adopted in the preset time, and selecting the total number of times each preferred knowledge text is adopted in the preset time as the application number of each preferred knowledge text; Acquiring feedback scores of each preferred knowledge text within preset time, selecting an average value of the feedback scores of each preferred knowledge text as an average feedback score of each preferred knowledge text, acquiring the front feedback times of each preferred knowledge text and the total feedback times of each preferred knowledge text within the preset time, and selecting a ratio of the front feedback times of each preferred knowledge text to the total feedback times of each preferred knowledge text as the front feedback rate of each preferred knowledge text, wherein the front feedback times of the preferred knowledge text are feedback times of the preferred knowledge text which are larger than the feedback times of the preset scores; Generating a comprehensive score of each preferred knowledge text according to the search frequency of each preferred knowledge text, the application frequency of each preferred knowledge text, the average feedback score of each preferred knowledge text, the front feedback rate of each preferred knowledge text and the comprehensive score model, selecting the preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring the influence range of a reference item and the solution of the reference item from the target knowledge text, fusing the characteristics of the influence range of the reference item, the characteristics of the solution of the reference item and the splicing characteristics of the development data of the IPTV item, generating the fusion characteristics of the IPTV item, inputting the fusion characteristics of the IPTV item into a large language model, and generating a summary report of the development data of the IPTV item through the large language model.
  5. 5. The summary report generating method according to claim 1, wherein after the generating of the comprehensive score of each preferred knowledge text based on the search frequency of each preferred knowledge text, the number of applications of each preferred knowledge text, the average feedback score of each preferred knowledge text, the positive feedback rate of each preferred knowledge text, and the comprehensive score model, selecting a preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring an influence range of a reference item and a solution of the reference item from the target knowledge text, fusing characteristics of the influence range of the reference item, characteristics of the solution of the reference item, and splice characteristics of development data of the IPTV item, generating fusion characteristics of the IPTV item, inputting the fusion characteristics of the IPTV item into a large language model, generating a summary report of the development data of the IPTV item by the large language model, the summary report generating method comprises: And creating a display window, and displaying a summary report of the development data of the IPTV project through the display window.
  6. 6. The summary report generating method according to claim 5, wherein after the creating of the display window, the summary report of the development data of the IPTV project is displayed through the display window, the summary report generating method comprises: and uploading a summary report of the development data of the IPTV project to the cloud platform.
  7. 7. The summary report generating method as recited in claim 1, wherein the IPTV item is a software development item of IPTV.
  8. 8. The summary report generating method as claimed in claim 1, wherein, The comprehensive scoring model is defined as follows: ; The higher the comprehensive score of the jth preferred knowledge text, the higher the probability of addressing the defect information of the IPTV project through the jth preferred knowledge text; representing the search frequency of the j-th preferred knowledge text; Representing the number of applications of the j-th preferred knowledge text; An average feedback score representing the j-th preferred knowledge text; representing the positive feedback rate of the j-th preferred knowledge text.
  9. 9. A summary reporting apparatus based on a large language model, applied to an electronic device, comprising: The acquisition module is used for acquiring the development data of the IPTV project, wherein the development data of the IPTV project comprises task information of the IPTV project, defect information of the IPTV project, progress information of the IPTV project and code submitting information of the IPTV project; The extraction module is used for respectively carrying out feature extraction on the task information of the IPTV project, the defect information of the IPTV project, the progress information of the IPTV project and the code submission information of the IPTV project by adopting a text feature extraction model to respectively obtain the features of the task information of the IPTV project, the defect information of the IPTV project, the progress information of the IPTV project and the code submission information of the IPTV project; The splicing module is used for splicing the characteristics of the task information of the IPTV project, the characteristics of the defect information of the IPTV project, the characteristics of the progress information of the IPTV project and the characteristics of the code submitting information of the IPTV project to obtain the splicing characteristics of the development data of the IPTV project; The retrieval module is used for retrieving the spliced features of the development data of the IPTV project in the knowledge base through the retrieval enhancement generation model to obtain a retrieval result, acquiring the similarity between each preset feature of the knowledge base and the spliced feature of the development data of the IPTV project in the retrieval result, selecting a plurality of preset features with the similarity higher than a preset value as a plurality of preferred features, and taking knowledge texts associated with all the preferred features as a plurality of preferred knowledge texts; The generating module is used for generating comprehensive scores of the preferred knowledge texts according to the search frequency of the preferred knowledge texts, the application times of the preferred knowledge texts, the average feedback score of the preferred knowledge texts, the front feedback rate of the preferred knowledge texts and the comprehensive score model, selecting the preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring the influence range of the reference item and the solution of the reference item from the target knowledge text, fusing the characteristics of the influence range of the reference item, the characteristics of the solution of the reference item and the splicing characteristics of the development data of the IPTV item, generating fusion characteristics of the IPTV item, inputting the fusion characteristics of the IPTV item into a large language model, and generating a summary report of the development data of the IPTV item through the large language model.
  10. 10. An electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the summary report generating method of any one of claims 1 to 8 when the computer program is executed.

Description

Summarization report generation method, device and equipment based on large language model Technical Field The application relates to the technical field of artificial intelligence and the technical field of project management, in particular to a summary report generation method, device and equipment based on a large language model. Background The chinese of IPTV is an interactive network television, and in order to improve the service capability and market competitiveness of IPTV, an IPTV project needs to be developed. In the development scenario of the IPTV project, project managers and technicians need to comprehensively understand the progress status of the IPTV project according to summary reports of the development data of the IPTV project. However, the summary report of the development data of the IPTV project depends on manual operation, which requires technicians to collect the development data from a plurality of decentralized channels, and also requires the technicians to spend a lot of time and time in the process of sorting out the summary report of the development data of the IPTV project according to the format required by project management, so that hysteresis occurs in the summary report of the development data of the IPTV project, which causes that project managers cannot judge the development risk of the IPTV project in time, and further affects the formulation efficiency of countermeasures. How to generate summary reports of development data of IPTV projects is a technical problem to be solved. Disclosure of Invention The embodiment of the application provides a summary report generation method, device and equipment based on a large language model, which are used for solving the technical problem of how to generate a summary report of development data of an IPTV project. In a first aspect, an embodiment of the present application provides a summary report method based on a large language model, which is applied to an electronic device, where the summary report generating method includes: Acquiring development data of an IPTV project, wherein the development data of the IPTV project comprises task information of the IPTV project, defect information of the IPTV project, progress information of the IPTV project and code submitting information of the IPTV project; Respectively extracting characteristics of task information of the IPTV project, defect information of the IPTV project, progress information of the IPTV project and code submission information of the IPTV project by adopting a text characteristic extraction model to respectively obtain the characteristics of the task information of the IPTV project, the defect information of the IPTV project, the progress information of the IPTV project and the code submission information of the IPTV project; Splicing the characteristics of the task information of the IPTV project, the characteristics of the defect information of the IPTV project, the characteristics of the progress information of the IPTV project and the characteristics of the code submitting information of the IPTV project to obtain the splicing characteristics of the development data of the IPTV project; Searching splicing features of the development data of the IPTV project in a knowledge base through a search enhancement generation model to obtain a search result, acquiring similarity between each preset feature of the knowledge base and the splicing features of the development data of the IPTV project in the search result, selecting a plurality of preset features with similarity higher than a preset value as a plurality of preferred features, and taking knowledge texts associated with all the preferred features as a plurality of preferred knowledge texts; Generating a comprehensive score of each preferred knowledge text according to the search frequency of each preferred knowledge text, the application frequency of each preferred knowledge text, the average feedback score of each preferred knowledge text, the front feedback rate of each preferred knowledge text and the comprehensive score model, selecting the preferred knowledge text with the highest comprehensive score as a target knowledge text, acquiring the influence range of a reference item and the solution of the reference item from the target knowledge text, fusing the characteristics of the influence range of the reference item, the characteristics of the solution of the reference item and the splicing characteristics of the development data of the IPTV item, generating the fusion characteristics of the IPTV item, inputting the fusion characteristics of the IPTV item into a large language model, and generating a summary report of the development data of the IPTV item through the large language model. In a possible implementation manner of the first aspect, the acquiring development data of the IPTV project, the summary report generating method includes: and sending a data acquisition request to the project managemen