Question And Answer Intention Recognition And Application Based On Deep Learning

Now more and more websites can provide information for users,but in terms of information bursting,how to provide users with the most useful information is a crucial concern.The general website question and answer mode is to let the user automatically select the type when the user submits a question.There are two problems in this process.First,this model will limit the data to this site and cannot provide more information and referenced data;Second,the users will not seriously and effectively identify the type of intent for the question sometimes.Therefore,the thesis mainly solves the above two problems,integrates information resources,provides users with more accurate and rich intent classification data.And implements the intent recognition system based on the deep learning model,which is very meaningful for improving the convenience of information.The main work of this thesis is as follows:(1)The data source is obtained through other sites,and the current popular travel websites are selected.This not only ensures the wide range of data sources,but also ensures the true validity of the data to provide rich data to meet the needs of users.(2)Intention recognition is a form of connection between question and answer sentences and users' needs.In the research of single intention recognition,Fast Text is chosen as the recognition model,which is fast and accurate.It is compared with the traditional word frequency distribution and common deep learning algorithm model,and the parameters of the algorithm model are analyzed and optimized.(3)For multi-intelligence identification data present in the system,this paper proposes a research method based on the combination of variable sliding window mechanism and Fast Text model.The sliding window mechanism fully considers the relationship between sentences,so as to improve the recognition effect of multi-intention problem,and at the same time,it can change the window size to be optimal.(4)The single intention and multi-intent model are combined the experimental results show that the hybrid model has better performance in the scenario of the intended recognition.In order to adapt to the operation of the intent identification system,supplemented by alarms,reports and other functions to ensure continuous and stable information for users.