| With the rapid development of the Internet in recent years,people are enjoying the convenience of big data while also facing the problems caused by the information explosion.Traditional search engines have been unable to meet the needs of people to quickly and accurately obtain information.However,the emergence of intelligent question answering systems has solved this problem well.In the field of game services,the number of game users nationwide reached 626 million till 2018,and the game industry revenue was 214.44 billion yuan,accounting for 23.6% of the global market revenue.Such a large-scale user will encounter various problems when playing games.They can only find web pages and posts arranged in chronological order by keywords in forums,posts,browsers,etc.,and the content is complex and almost no satisfactory answer can be found.To solve those problems,an intelligent question answering system,which is based on deep learning in the field of game corpora is constructed in this article.This system can accept the user's natural language and return to the game user a precise answer after information retrieval and question and answer matching,which greatly improves the user's gaming experience.It is of great help to increase game activity and reduce operator costs,so there is a lot of market development space and application value.The main research contents of this article are as follows:(1)There is currently no public data set in the limited field.This paper collects game knowledge question and answering corpus through domain websites,Zhihu and Sogou question and answering platform,et al.And obtains game question and answer corpus through a crawler program.After removing unnecessary noise after data cleaning,we store data in the database to complete the establishment of a limited domain corpus knowledge base.(2)In the information retrieval section,this paper firstly analyzes the principle and structure of the recurrent neural network(RNN),the bidirectional long-term and short-term memory network(BLSTM),the hybrid network(BLSTM / CNN),and the hybrid neural network incorporating the attention mechanism.And we prove that the LSTM / CNN hybrid neural network integrated with attention mechanism is superior in information retrieval through experiments and improve the accuracy of the question answering system.The specific process of information retrieval is that the system receives the text sent by the user,and then inputs the text to the BLSTM / CNN mixed model integrated with attention mechanism,finally outputs the final result after information retrieval and cosine similarity calculation.(3)This article designs and implements a game corpus intelligent question answering system based on the above hybrid model.The system's deep learning algorithm uses Python language and Google's Tensorflow deep learning framework.The system chooses Spring Boot lightweight framework in the background,the data persistence layer uses MySQL database,and the data access layer uses the enterprise-level MyBatis framework.The user submits the question through the web interface.After receiving the natural language text in the background,the system returns the corresponding answer to the user through searching and matching in the game corpus knowledge base.Then the system displays the results on the web page.This paper not only verifies the hybrid model incorporating the attention mechanism,but also uses this model to complete the construction of a game corpus intelligent question answering system. |
