Intelligent question and answer method, device and equipment for knowledge in power production field and medium

Abstract

The invention relates to the technical field of artificial intelligence, in particular to an intelligent question answering method and device for knowledge in the power production field, equipment and a medium, and the method comprises the steps: carrying out semantic partitioning and vectorization embedding on a knowledge base document, forming a data set and a vector set in the power production field, and obtaining a knowledge problem about the power production field inquired by a user; respectively executing vector retrieval and keyword retrieval on the data set and the vector set according to a knowledge problem about the power production field inquired by a user to obtain a multi-path retrieval recall result; and fusing multi-path retrieval recall results to obtain similarity scores Top T vector document blocks to form a candidate set, inputting the vectorized knowledge questions and the vector document block candidate set into a large model, and generating question and answer results of the knowledge questions by the large model. Therefore, the problems that diversified information resources are difficult to fully utilize due to the adoption of a single retrieval path in related technologies, and processing limitation on timeliness and multi-modal data is caused are solved.

Claims

1. An intelligent question-answering method for knowledge in the field of power production, characterized in that: The following steps are involved: Perform semantic segmentation and vector embedding on knowledge base documents to form data sets and vector sets in the field of power production, and obtain user queries on knowledge related to the field of power production; performing vector search and keyword search on the data set and the vector set respectively according to the user's query on the knowledge question in the field of power production, and obtaining a multi-way search recall result; The multi-way retrieval recall results are integrated to obtain a vector document block candidate set, the vectorized knowledge question and the vector document block candidate set are input into a large model, and the large model generates a question-answering result for the knowledge question.

2. The intelligent question-answering method for power production domain knowledge according to claim 1 is characterized in that: The semantic segmentation and vectorization embedding of the knowledge base documents to form a data set and vector set in the field of power production include: Semantically segment the knowledge base documents into blocks to generate a set of document blocks; The document block set is embedded in a text embedding model, and the text embedding model vectorizes the divided document blocks to form a data set and a vector set in the field of power production.

3. The intelligent question-answering method for power production domain knowledge according to claim 2, characterized in that: The step of semantically segmenting the knowledge base documents to generate a document block set includes: Divide the document into chunks based on the sentence or paragraph units in the knowledge base document; The segmented knowledge base documents are input into a text embedding model, which converts each sentence or paragraph into a target dimensional space to generate a vector set; Calculate the cosine similarity between each sentence vector block or paragraph unit vector block in turn; The vector blocks are grouped into document blocks according to the cosine similarity until all vector blocks are allocated.

4. The intelligent question-answering method for power production domain knowledge according to claim 3 is characterized in that: The step of performing vector search and keyword search on the data set and the vector set respectively according to the user's query on the knowledge question in the field of power production, and obtaining a multi-way search recall result, includes: Mapping document blocks in the new document block set to a multidimensional vector space to generate vector document blocks; Mapping knowledge problems in the field of power production into a multi-dimensional vector space to generate problem vectors; Calculating the cosine similarity between the question vector and the vector document block; The document blocks are sorted in descending order according to their cosine similarity to generate multi-way retrieval recall results.

5. The intelligent question-answering method for power production domain knowledge according to claim 4 is characterized in that: The step of performing vector search and keyword search on the data set and the vector set respectively according to the user's query on the knowledge question in the field of power production, and obtaining a multi-way search recall result, includes: Calculate the keyword search similarity score of the question vector and the vector document block based on word frequency, inverse document frequency, and document length; The document blocks are sorted in descending order according to their keyword retrieval similarity scores to generate multi-way retrieval recall results.

6. The intelligent question-answering method for power production domain knowledge according to claim 5, characterized in that: The vector document block candidate set is obtained by integrating the multi-way retrieval recall results, including: Calculate the fusion similarity based on the ranking position of the multi-retrieval recall results and the differentiation weight coefficient; The vector document block candidate set is generated by filtering and fusing the multi-way retrieval recall results according to the fusion similarity.

7. An intelligent question-answering device for knowledge in the field of power production, characterized in that: include: The acquisition module is used to perform semantic segmentation and vector embedding on the knowledge base documents to form a data set and vector set in the field of power production, and obtain user queries on the knowledge questions in the field of power production; a generation module, configured to perform vector search and keyword search on the data set and the vector set respectively according to the user's query on the knowledge question in the field of power production, and obtain a multi-way search recall result; A fusion module is used to fuse the multi-channel retrieval recall results to obtain a vector document block candidate set, input the vectorized knowledge questions and the vector document block candidate set into a large model, and the large model outputs the question-answering results of the knowledge questions.

8. An electronic device, characterized in that: include: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the intelligent question-answering method for knowledge in the field of electric power production as described in any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The program is executed by a processor to implement the intelligent question-answering method for knowledge in the field of power production as described in any one of claims 1 to 6.

10. A computer program product, characterized in that It includes a computer program, which, when executed by a processor, is used to implement the intelligent question-answering method for knowledge in the field of power production as described in any one of claims 1 to 6.

Images