Mongolian multi-hop question and answer method based on three-channel cognitive map and map attention network

Abstract

The invention discloses a Mongolian multi-hop question and answer method based on a three-channel cognitive map and a map attention network. The method comprises the following steps: translating a Mongolian query question into a Chinese query question by machine translation; performing part-of-speech tagging on the Chinese query question after sentence segmentation and word segmentation; converting the Chinese question statement subjected to part-of-speech tagging into a Chinese query statement, and inputting the Chinese query statement into a cognitive map server; and allowing the cognitive map server to simulate a cognitive system of human beings in cognition, using three channels for parallel computing, finally, through normalization, query sorting and query selection, giving an answer with the maximum probability , and returning a result . Question and answer of different requirements are performed by means of respective advantages of the three channels, the accuracy and speed of a question and answer system in natural language processing are greatly improved, meanwhile, the channels 1, 2 and 3 can be reused in translation and question and answer stages, the speed is nearly doubled, and the question and answer quality is integrally improved.

Description

technical field [0001] The invention belongs to the cross-technical field of question answering system and cognitive science in natural language processing (NLP), and in particular relates to a Mongolian multi-hop question answering method based on a three-channel cognitive graph and graph attention network. Background technique [0002] Machine reading comprehension and question answering systems have long been considered one of the core issues of natural language understanding (NLU). With the rise of models such as BERT, a major breakthrough has been made in simple reading comprehension tasks of single paragraphs; but in the "multi-hop", The accuracy rate in the "complex" situation did not reach human level. [0003] Question Answering (QA) is an important research field of Natural Language Processing (NLP). In this field, researchers aim to build a system that can automatically provide answers to questions posed by humans in "natural language." [0004] Different from t...

AI Technical Summary

Problems solved by technology

[0011] In order to overcome the shortcomings of the above-mentioned prior art, the object of the present invention is to provide a Mongolian multi-hop question answering method based on a three-channel cognitive map and graph attention network, in order to solve the problem of Mongolian knowledge base and knowledge in the current Mongolian question answering scene. For problems lacking in maps, realize the integration of fast query of simple questions and inference query of complex questions, improve the adaptability and accuracy of Mongolian question and answer, and provide a better question and answer system method

Images