Research On Automatic Question Answering Technology Based On Knowledge Graph

Knowledge graph is an effective way to organize and manage massive amounts of data.The automatic question answering system based on knowledge graph is a technology that allows users to give questions in the form of natural language,analyze and process the questions,and obtain accurate answers by querying the knowledge graph.There are still many challenges in achieving an efficient and accurate question and answer system.For example,there is a huge semantic gap between natural language forms and structured knowledge representations.For example,a question can be understood in many different ways.The way to make the answer to the question is not accurate enough.At present,the main idea of implementing question and answering in knowledge graph is to decompose question and answering into two sub-tasks,entity extraction and relationship detection.For these two tasks,this paper proposes entity extraction algorithm based on LSTM model and WRL-BiLSTM relation detection algorithm based on deep learning.The details are as follows.(1)The entity extraction algorithm based on the LSTM model aims to determine whether each word in the question is part of the entity through the LSTM model.The result of the final output of the question through the word representation layer,the feature extraction layer,and the conclusion inference layer is the category information of each word in the question.The result of the model output is further extended by a heuristic method to obtain the entity in the question.Finally,the experimental comparison on the SimpleQuestions dataset shows that the algorithm can significantly improve the recall rate of the entity extraction task.(2)WRL-BiLSTM relation detection algorithm based on deep learning,by combining semantic level matching and text level matching between questions and relationships,the model can capture high-level semantic information between questions and relationships while capturing semantic details.The change on.The experimental results on the SimpleQuestions dataset show that the features extracted based on the above two matching modes help to improve the accuracy of the relationship detection task.Finally,the results of the integrated entity extraction task and the relationship detection task are queried on the Freebase knowledge graph to obtain the final question answer prediction result,and compared with the existing question and answer model,the accuracy of the result is significantly improved.The paper has 17 pictures,14 tables and 61 references.