Design And Implementation Of Intelligent QA Reasoning System Based On Natural Language Entity Relationship

Nowadays,with the acceleration of informatization construction in various industries and the expansion of industrial scale,people who obtain more cutting-edge information more accurately and efficiently in order to broaden their horizons and increase awareness,will use intelligent question and answer systems for queries.Intelligent question answering system is an information retrieval system based on artificial intelligence that is more powerful and efficient than traditional search engines.It is also an important form of interaction between users using natural language and computers.A good intelligent question answering system requires the participation of a rich underlying corpus in order to provide reasoning decision support,so how to organize the relevant information needed for decision-making from the corpus to help the system better complete the reasoning process is particularly important.There are two difficulties involved here: how to extract semantic-rich entity connection relationship data based on the inherent characteristics and structure of the corpus,and how to perform intelligent and efficient reasoning based on these data to answer user questions.Therefore,automatically constructing entity relationship data related to user questions from the underlying corpus and obtaining answers from it through intelligent and efficient reasoning methods becomes very valuable for research.This thesis will use the Chinese Wikipedia as the basic corpus to conduct research on the key technologies of intelligent question answering inference methods based on the connection of natural language entities,and hope that it will be extended to some specific corpora in the future to achieve corresponding results.The research work mainly includes designing Chinese entity extraction algorithms,deriving key entity information from problem description sentences,and using corpus to construct natural language entity connection relationships to complete the construction of reasoning decision data,and finally combining BERT pre-training language models and graphs Neural network architecture to design efficient intelligent question answering inference algorithms.Experiments show that the entire system has strong reasoning and analysis capabilities,and can give reasonable and correct answers based on the user's complex question text.This article also develops a set of available interactive interface for intelligent question answering system based on this,which has good interactive capabilities and user experience.