Research On Telephone Text Classification Model And Application Based On Deep Learning

In the current era,the development of China's tourism e-commerce industry tends to be mature and stable.Major enterprises pay more and more attention to user services in order to occupy a place in the industry and attract and maintain the target users of the enterprise.User call is an important and direct way for companies to understand users and discover problems.By classifying the telephone text of users' call records,the travel company can efficiently collect important information carried by the telephone text,analyze the users' needs,and discover problems in the product or the enterprise in time.The classification of telephone text is of great significance in enterprise decision-making such as service improvement,product marketing,user maintenance and enterprise competition.This paper uses OTA(Online Travel Agency)user telephone text as the data foundation,aiming to build a telephone text classification model.Based on the introduction and research of telephone text classification,the related process of text classification and related methods of current text classification at home and abroad are described,and the advantages and limitations of existing text classification methods are compared and analyzed.Then,after observing and analyzing the characteristics of the telephone text,the classification system is optimized based on the life cycle theory,and the categories of telephone texts are sorted and optimized to increase the degree of text differentiation.At the same time,a multi-path telephone text classification model with multi-feature fusion is proposed,and text representation is performed from multiple perspectives.The features are extracted by convolutional neural network,long-term and short-term memory network and attention mechanism,and input into Softmax classifier to realize classification.Finally,based on the user telephone text data of an OTA enterprise,the model experiment study is carried out,and the established model is evaluated by using the accuracy index and other evaluation indexes,and the feasibility and practical significance of the model applied to the telephone text classification are analyzed.