Intelligent question answering method and device

Abstract

The present invention provides an intelligent question answering method and device. The method comprises the steps of performing word segmentation processing on the text to be solved, and determininga context for semantic similarity judgment according to the word segmentation result of the problem to be solved; collecting a certain number of frequently asked questions according to the context; carrying out word segmentation processing on the text of all the frequently asked questions, and building a context map according to the word segmentation results of all the frequently asked questions;for any of that frequently ask questions, calculating the semantic similarity between the frequently ask questions and the questions to be answered according to the word segmentation result of the frequently asked questions, the word segmentation result of the questions to be answered and the context map; taking candidate answers corresponding to frequently asked questions having the highest similarity as answers corresponding to the questions to be answered. The embodiment of the invention can more accurately analyze the questions to be answered and provide the answers.

Description

technical field [0001] The present invention relates to the technical field of natural language processing, and more specifically, to an intelligent question answering method and device. Background technique [0002] In the question answering system, the push of general chat answers is highly random. But in the field of professional applications, the content of the reply needs to be precise. The study of comparing the semantics of "user questions" with computer recognition and existing sentences in the sentence database is called sentence similarity research. As a key issue in natural language processing, it has always been a research hotspot and difficulty. In addition to mining the inter-word relationship of the sentence itself and calculating the sentence similarity by overlapping degrees (such as relying on the WordNet architecture and relying on the HowNet architecture and corpus), the research on sentence similarity has also begun to develop feature extraction based ...

AI Technical Summary

Problems solved by technology

This method considers the relevant elements of semantic information, but the construction is relatively complicated, such as semantic similarity calculation based on ontology, etc.

The third method, based on the neural network corpus training feature extraction method, has also been vigorously developed in recent years, such as the study of sentence semantic similarity calculation based on Word2vec, which relies on the quality and quantity of corpus, pays attention to feature extraction, and ignores sentence meaning. understanding, cannot realize the real mining of semantics

This method draws on the relative definition of the membership degree to construct the fuzzy importance of words in the context, which improves the similarity of the word's sentence meaning, but it is insufficient in the overall sentence similarity of the sentence.

Images