Design And Implementation Of Intelligent Q&A System Based On Chinese Knowledge Graph

With the continuous development of artificial intelligence technology,the traditional methods for knowledge acquisition which are based on search engines are increasingly difficult to meet the needs of people who obtain information from the Internet.The returned information is so complex that the users need to spend a lot of time for the correct answer.Meanwhile,The Intelligent Question Answering System is taken more seriously for its able to accurately capture the user's search intention,to understand their natural language question and to return the answer directly to them.The rapid development of Knowledge Graph,which have provided a high-quality source of knowledge for the implementation of Intelligent Question Answering System,has directly promoted the development of the system in industry fields(such as customer service and medical care).This thesis designed an Intelligent Question Answering Framework which is based on Chinese Knowledge Graph.With the functional modules including data processing?entity identification,and attribute connectivity,the framework uses knowledge of generic Chinese Knowledge Graph as a source of knowledge.It can accurately identify the subject of inquiry from the user's diversely expressed questions and find answers from the knowledge graph.The main work of this thesis is as following:1.In response to the problem that the context of traditional semantic representation is insufficient,this thesis adopts word embedding to extract the semantic features of the user's natural language question as the model input of the entity recognition and attribute linking module.2.Based on the diversity and ambiguity of user expressions in question sentences,Bi-LSTM(Bidirectional Long-Short Memory)model was introduced.Using Bi-LSTM to extract semantic features and annotate entities can better identify multiple Words such as ambiguity so that can improve the accuracy of entity recognition.3.For the problem of insufficient deep-level semantic feature extraction,this thesis introduces the attention mechanism in the CNN(Convolutional Neural Networks)model which makes the CNN model a better mine of the semantic relationship between natural language question words and attribute words.Experimental results show that the introduction of Bi-LSTM in the entity recognition algorithm and the integration of attention mechanism in the attribute linking algorithm can improve the accuracy of answering questions,and it verified on an open data set.In this thesis,an Intelligent Question Answering System based on Chinese knowledge graph is implemented according to the above algorithm.It runs well in the experimental testing environment and can accurately answer natural language questions.At the same time,we also tested the function and performance of the designed system in the actual environment.The result shows that the system has a good sense of stability,which proves that the above algorithm is feasible.