Research And Implementation Of Automatic Question Answering System Of Incomplete Information Game Based On Knowledge Graph

With the successful holding of the 2008 world mind sports games in Beijing,intellectual sports are flourishing in China.In the intellectual sports,incomplete information game accounts for a large proportion.People's enthusiasm for learning incomplete information game intelligence project is high.However,in the field of incomplete information game,the rules of various competitive events are complicated,and there are many enthusiasts and few experts,which cannot meet the needs of the masses to learn and solve problems.And machine game is always the touchstone of artificial intelligence technology.Nowadays,most of the intelligent decision models of incomplete information game based on artificial intelligence technology are black box models.How to implement the decision interpretability is also one of the development directions of game decision field.In order to meet the learning needs of incomplete information game enthusiasts and promote the development of intelligent decision interpretability in incomplete information game,this paper proposes and implements an automatic question-answering system based on the knowledge graph of incomplete information game field.The completed work can be divided into the following parts:1.The knowledge graph of incomplete information game is designed and constructed.Based on the seven-step method of domain ontology construction proposed by Stanford University School of Medicine,this paper combines unsupervised learning algorithm to automatically mine domain specific terms,and designs the definition of knowledge pattern in this field.This paper establishes a corpus in the field of incomplete information game.Based on the BiLSTM-CRF model,a named entity recognition model of incomplete information game field is proposed.Based on the BiLSTM-Attention model,a relationship extraction model of incomplete information game field is proposed.In this paper,the corpus is extracted by the model,and the knowledge is audited by the combination of manual audit and rule audit,and the accurate knowledge is stored in the Neo4j database in the form of triple.2.This paper designs an automatic question answering based on incomplete information game knowledge graph.According to the order of named entity identification,entity link,intention identification and answer retrieval,this module processes the user's questions and implements automatic question answering.BiLSTM-Attention model is used to implement the intention identification of incomplete information game domain.By analyzing the entities and intentions extracted from the questions,answer retrieval generates the Cypher queries used in the Neo4j database and queries the answers in the database.3.In this paper,an assistant intelligent question-answering system based on incomplete information game teaching is implemented.This paper implements the front-end of the question-answering system based on Android,introduces the speech recognition module of iFLYTEK to implement the multi-form question input of voice and text.In this paper,the back-end of question-answering system is implemented based on Flask framework of Python,and the tasks of problem analysis,answer retrieval and generation are completed.Through testing,the system can answer most questions in the field of incomplete information game,and can assist enthusiasts in incomplete information game learning.