Retrieval type intelligent question-answering method and device based on multi-model fusion

Abstract

The invention discloses a retrieval type intelligent question-answering method and device based on multi-model fusion. The method comprises the steps: carrying out the text preprocessing of a user question sentence, carrying out the word segmentation and named entity recognition, and obtaining a word segmentation result of the question sentence and an entity mark of the question sentence; and enabling the word segmentation result and the entity mark to pass through a rule classifier, matching rules in a rule base, and returning the standard questions corresponding to the rules and answers to the standard questions to the user. The effect of the model is improved; in the rule classifier, an automatic rule generation method is used, a large amount of manpower expenditure is saved. Meanwhile,almost 100% of accuracy can be achieved for questions in a knowledge base, and the use feeling degree of customers is improved; when the corpus is enough, the effect and generalization performance ofthe model are improved, and the accuracy of intention recognition is improved; the operation efficiency of the algorithm model is optimized, and the operation speed of the algorithm is improved.

Description

technical field [0001] The invention belongs to the field of intelligent question answering, and in particular relates to a retrieval type intelligent question answering method and device based on multi-model fusion. Background technique [0002] Existing intelligent question answering is mainly based on information retrieval or semantic understanding technology to find answers from a large number of candidate sets. When a user asks a question, the question will be matched in the index database; firstly, the keyword search will be roughly sorted, and some question-answer pairs that may match the answer will be recalled; then, the fine sorting calculation will be performed through semantics and other richer algorithms, and the best one will be returned a result. Flowchart such as Figure 5 shown. [0003] A commonly used method of keyword rough sorting is to use Elasticsearch for word segmentation query, and query the knowledge base through the word segmentation results of ...

AI Technical Summary

Problems solved by technology

[0005] Retrieval question and answer is easy to understand and easy to understand, but there are many disadvantages, such as: (1) Retrieval question and answer is very dependent on a pre-defined database, requiring a large number of high-quality question and answer pairs or knowledge points; (2) Keyword rough sorting Retrieval determines the upper limit of the accuracy of the algorithm. Although the retrieval speed is fast using Elasticsearch, it does not consider semantic information

[0006] Traditional search-based question-and-answer requires a large amount of high-quality corpus and knowledge base. When there is a large amount of knowledge base data, there is a problem of low recognition accuracy.

Images