Research On Key Technologies Of Intelligent Chat Robot Based On Deep Learning

With the rapid development of deep learning in the field of natural language processing,research on techniques such as text analysis,information retrieval,part-of-speech tagging and question-answering systems has become more mature.In particular,chat robot technology is becoming mature with the expansion of the portable mobile device market,and is widely used in customer service,finance,medical care,education and life services.Because the chatter based on Chinese corpus has the excellent characteristics of open interaction,this technology will greatly improve people's lifestyle and become a technological growth point in the new era.At present,when studying chat robots,scholars usually use the Seq2Seq model based on deep learning to perform word vector conversion and model training on the constructed corpus.This paper analyzes and studies the common problems and defects that are easy to appear in the crucial stages of the chat robot's development process,and proposes improvement methods.Then this paper analyzes the problem that the current sports app is ubiquitous and has insufficient interaction with the user,and applies the improved chat robot model to the sports app to enhance the interaction experience with the user.The specific research work has the following aspects:(1)An improved algorithm is proposed for the traditional construction of Chinese-specific corpus with limited number of corpora and weak dialogues.The corpus constructed by the new method is not only richer than the corpus constructed by the existing general methods,but also the correlation between the dialogues is stronger.(2)In order to improve the training effect of Chinese word vector in training,combined with the two optimization algorithms of Word2Vec,the Chinese corpus is trained by the model of combined algorithm,which makes the expression of Chinese word vector more accurate.(3)When the traditional chat robot Seq2Seq model is used to train the dialogue corpus,it will produce too safe,normal and repeated answers,which affects the user interaction experience.This paper combines the idea of mutual information to improve the objective function of the model,which can produce richer and more diverse answers,giving users a good experience.(4)Aiming at the problem that the current smart terminal App is ubiquitous and the user interaction experience is poor,the mobile chat assistant that combines the chat robot and the sports app is designed and implemented,so that the user can communicate with the app while exercising,and this enhances the freshness and interaction of the user experience.