Search

CN-121997909-A - Knowledge editing and updating method, system, equipment and medium for cross-language insurance question and answer

CN121997909ACN 121997909 ACN121997909 ACN 121997909ACN-121997909-A

Abstract

The application relates to the technical field of intelligent decision making and discloses a method, a system, equipment and a medium for editing and updating knowledge of cross-language insurance questions and answers, wherein models of different languages finish unified storage of documents of different languages based on a common semantic coordinate system; when the insurance company updates the policy or the product description, the knowledge editing function triggers the positioning module to identify and position the content change points, when the Chinese model generates the knowledge correction gradient signal, the knowledge editing function positions the content change points corresponding to the knowledge correction gradient signal, and the shared editing space is used for completing one-time modification and global synchronization cross-language synchronous updating, and the cross-language consistency verification is carried out on the content after the synchronous updating, and the necessary correction optimization is completed. The application can be applied to the development of business systems such as financial science and technology, medical health care and the like, avoids the redundant operation of modifying the traditional multilingual one by one, and reduces the repeated labor cost of multilingual teams.

Inventors

  • ZHANG NAN

Assignees

  • 平安科技（深圳）有限公司

Dates

Publication Date
20260508
Application Date
20260115

Claims (10)

  1. 1. A knowledge editing and updating method for cross-language insurance questions and answers is characterized by comprising the following steps: the models of different languages finish unified storage of documents of different languages based on a common semantic coordinate system; when the insurance company updates the policy or the product description, the knowledge editing function triggers the positioning module to identify and position the content change point; When the Chinese model generates a knowledge correction gradient signal, a content change point corresponding to the knowledge correction gradient signal is positioned through a knowledge editing function, and once-modified and globally-synchronized cross-language synchronous updating is completed through a shared editing space; performing cross-language consistency verification on the content after synchronous updating, and completing necessary correction optimization; And after all language versions are updated, executing a safe output release process.
  2. 2. The method for editing and updating knowledge of cross-language insurance questions and answers according to claim 1, wherein the step of completing unified storage of documents in different languages based on a common semantic coordinate system by using the models in different languages comprises the following steps: Constructing a common semantic coordinate system of different language models; through training a cross-language embedding model, text in different languages is encoded into vector representation, so that synonymous concepts represent semantic aggregation characteristics in a high-dimensional space; the knowledge retrieval layer stores the vectorized knowledge documents of all languages through a unified vector index base.
  3. 3. The method for updating knowledge editorial of cross-language insurance questions and answers as claimed in claim 1, wherein the step of triggering the location module by the knowledge editorial function to identify and locate the content change point when the insurance company updates the policy or the product description comprises: When an insurance company updates a policy or a product description, a document monitoring mechanism and a semantic difference detection technology are combined to realize accurate perception of content change; And calculating the vector distance of the documents before and after updating, triggering the knowledge node to update if the change of the vector distance exceeds a preset threshold, and generating a structured editing task.
  4. 4. The method for editing and updating knowledge of cross-language insurance questions and answers as claimed in claim 1, wherein the step of realizing accurate perception of content change by combining a document monitoring mechanism and a semantic difference detection technology when an insurance company updates a policy or a product description comprises: when the insurance company updates the policy or the product description, the content change of the new and old documents is compared by combining the newly uploaded clause file in the RAG knowledge base, so that the accurate perception of the content change is realized.
  5. 5. The method for updating knowledge of cross-language insurance question and answer according to claim 1, wherein when the Chinese model generates the knowledge correction gradient signal, the step of locating the content change point corresponding to the knowledge correction gradient signal through the knowledge editing function and completing the cross-language synchronous updating of one-time modification and global synchronization through the shared editing space comprises the following steps: After a knowledge correction gradient signal is generated in a Chinese model, the corresponding model update is corrected and presented in a mathematical form of a low-rank matrix, namely a lightweight parameter patch; mapping the parameter patches to a shared editing space, transmitting the parameter patches to each target language model through a language mapping matrix, and simultaneously ensuring that answer meanings of different languages are kept highly uniform by restricting the consistency of output semantics of the multiple language models.
  6. 6. The method for editing and updating knowledge of cross-language insurance questions and answers as claimed in claim 1, wherein the step of performing cross-language consistency verification on the content updated synchronously and completing necessary correction optimization comprises the steps of: calculating the similarity of multi-language output through a semantic consistency checking system, and verifying the semantic correctness of the content after synchronous updating; Retrieving corresponding treaty from the rule knowledge base of each country, and checking the consistency of the content and the local law; if uncertainty exists in the compliance verification process, a compliance difference prompt is generated, and a handover method staff performs manual confirmation.
  7. 7. The method for editing and updating knowledge of cross-language insurance questions and answers according to claim 1, wherein the step of executing the secure output issuing process after all language versions are updated comprises: after all language versions are updated, generating a knowledge updating record containing version identifiers; and synchronizing the knowledge updating record to the RAG question-answering system of each country, outputting an automatic report, and listing the modification content, the influence range and the manual auditing state.
  8. 8. A knowledge editing and updating system for cross-language insurance questions and answers, comprising: The storage module is used for completing unified storage of the documents in different languages based on a common semantic coordinate system by using the models in different languages; the positioning module is used for triggering the positioning module through a knowledge editing function to identify and position the content change point when the insurance company updates the policy or the product description; the updating module is used for positioning content change points corresponding to the knowledge correction gradient signals through the knowledge editing function when the Chinese model generates the knowledge correction gradient signals, and completing one-time modification and global synchronization cross-language synchronization updating through the shared editing space; the verification module is used for verifying cross-language consistency of the synchronously updated content and completing necessary correction optimization; And the output module is used for executing the safe output release flow after all the language versions are updated.
  9. 9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the knowledge editing update method of a cross-language insurance question and answer as claimed in any of claims 1 to 7.
  10. 10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the steps of the knowledge editing and updating method of cross-language insurance questions and answers as claimed in any one of claims 1 to 7.

Description

Knowledge editing and updating method, system, equipment and medium for cross-language insurance question and answer Technical Field The invention relates to the technical field of intelligent decision making, in particular to a method, a system, equipment and a medium for editing and updating knowledge of cross-language insurance questions and answers. Background In current insurance question-answering systems, knowledge updating of multilingual models is mainly achieved through full-scale retraining or independent localized fine tuning. For example, after the term expansion of "earthquake risk" is performed on the chinese insurance question-answering model, the thai, vietnamese and indonesia versions need to be retrained respectively to ensure the content synchronization. However, this approach has the following drawbacks: The method has the advantages that the updating efficiency is low, repeated labor is caused by independent training of each language, knowledge updating cannot be synchronized in real time, often a period of several weeks is needed, cross-language deviation is serious, semantic drift exists on the understanding of insurance clause details (such as regions, pay proportion and disclaimed areas) of different language models, and especially, after machine translation fine adjustment, semantic alignment is insufficient, and compliance risks are easy to occur. Disclosure of Invention The invention provides a knowledge editing and updating method, a system, computer equipment and a medium for cross-language insurance questions and answers, which are used for solving the technical problems of low updating efficiency and serious cross-language deviation in the prior art. In a first aspect, a method for editing and updating knowledge of a cross-language insurance question and answer is provided, including: the models of different languages finish unified storage of documents of different languages based on a common semantic coordinate system; when the insurance company updates the policy or the product description, the knowledge editing function triggers the positioning module to identify and position the content change point; When the Chinese model generates a knowledge correction gradient signal, a content change point corresponding to the knowledge correction gradient signal is positioned through a knowledge editing function, and once-modified and globally-synchronized cross-language synchronous updating is completed through a shared editing space; performing cross-language consistency verification on the content after synchronous updating, and completing necessary correction optimization; And after all language versions are updated, executing a safe output release process. In a second aspect, a knowledge editing and updating system for cross-language insurance question and answer is provided, including: The storage module is used for completing unified storage of the documents in different languages based on a common semantic coordinate system by using the models in different languages; the positioning module is used for triggering the positioning module through a knowledge editing function to identify and position the content change point when the insurance company updates the policy or the product description; the updating module is used for positioning content change points corresponding to the knowledge correction gradient signals through the knowledge editing function when the Chinese model generates the knowledge correction gradient signals, and completing one-time modification and global synchronization cross-language synchronization updating through the shared editing space; the verification module is used for verifying cross-language consistency of the synchronously updated content and completing necessary correction optimization; And the output module is used for executing the safe output release flow after all the language versions are updated. In a third aspect, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, the processor implementing the steps of the above-described knowledge editing and updating method for cross-language insurance questions and answers when the computer program is executed by the processor. In a fourth aspect, a computer readable storage medium is provided, the computer readable storage medium storing a computer program, which when executed by a processor, implements the steps of the above-described knowledge editing and updating method for cross-language insurance questions and answers. In the scheme realized by the knowledge editing and updating method, the system, the computer equipment and the storage medium of the cross-language insurance question and answer, unified storage of different language documents can be completed based on a common semantic coordinate system through models of different languages, when insurance companies update policies or product descript