Text Style Transfer Model Based On Keyword And Syntax Tree

Text style transfer is a cutting-edge and important subdivision field in the field of natural language processing.On the basis of previous research,this paper finds the deficiencies of previous models through comparison,and then finds the research goal of this paper,that is,the existing models cannot achieve the balance between text content and style retention,and the existing models often can only target one This paper proposes a text style transfer model based on keywords and syntax trees,which not only solves the above two problems of existing models This is a problem,and try to apply it to the chatbot task related to natural language processing,thus realizing the exploration from theory to industrial landing.Inspired by research in the field of natural language,this paper proposes a text style transfer model based on keywords and syntax trees.From the perspective of words and word order,the overall framework of the model is designed as a two-stage processing task,that is,finding the target first A list of keywords and words that may be expressed in the text,and then sort these word lists through the fusion syntax tree,and output the target text that retains both semantics and grammar,and designs and conducts experiments on two types of data,parallel corpus and non-parallel corpus In addition,this paper uses the mature evaluation indicators of text translation to verify the model experimentally,which confirms the effectiveness of the model.The core contributions of this paper mainly include the following aspects:(1)This paper implements a model based on keywords and syntax trees,which not only preserves the semantics of the original text,but also realizes the controllable conversion of text styles,and the generated texts are more readable.It solves the problem that the previous model cannot achieve a balance between text content and style retention.(2).This paper proposes a predictive ranking framework.It can not only realize the processing of supervised learning tasks and unsupervised learning tasks.By replacing the predictor module,supervised tasks and unsupervised tasks can be optimized in a targeted manner,which solves the problem that existing models often only target one type of corpus.(3).Based on the above model,this paper develops a chatbot APP that can customize the language style of the robot.By specifying the chat style of the robot,the chat robot can have a unified language style when replying to the user.It solves the problem that the text style returned by the trained chatbot to the user is not uniform due to the diversity and uncertainty of the training corpus in the chatbot task.