Research on fast construction of question answering system based on target graph

Aiming at the problems of chaotic structure and lack of visual query methods in small sample domain data, this paper proposes a fast construction framework of question answering (QA) system based on target graph. Firstly, we construct a knowledge graph by extracting entities, properties and relationships from open source datasets. Secondly, the method based on dictionary technology and Aho-Corasick (AC) automaton is used to realize fast entity recognition and question word extraction. Then a problem base is established for problem template matching. Finally, we use query statements to retrieve answers in the graph database. In the military field, experiments show that through the framework proposed in this paper, a more comprehensive target graph can be quickly constructed, on this basis, accurate QA can be realized.