Research And Implementation Of Long Comment Generation System For Automotive Field

With the popularization and development of Internet technology,the Internet has been integrated into many aspects of production and life.When it comes to automobiles,potential consumers often learn about automobile related information through the Internet.Long comments that are fluent,vivid and rich in car description information can enhance consumers’ desire to buy cars.However,it takes a lot of manpower to write comments manually.Automatically generating car comments becomes a task with development prospects and practical significance.According to the above industrial background,this thesis studies an encoder-decoder long comments generation model based on Conditional Variational Autoencoder,which improves the problem of incoherent semantics and lack of diverse expressions in long text generation.To make the generated comments semantically coherent and logically smooth,this thesis designs a clause content planning module,which uses high-level semantic information to guide the generated content of clauses and the logical structure between clauses.The latent variables are divided into layers,combined with the hierarchical decoding structure,the diversity is injected at different levels in the generation stage.At the same time,the idea of diverse beam search is introduced into the word decoder,so that the decoder can generate multiple comments and guarantee the difference between generated texts at the same time.Through the comparative experiment with the baseline model on the automobile dataset,it is verified that the comments generation model proposed in this thesis has a better generation effect in terms of sentence fluency and diversity.Due to the lack of public datasets in the automotive vertical field,this thesis constructs a knowledge graph in the automotive domain rich in attributes,parameters and other descriptive information with the CRF plus Bootstrap method for knowledge extraction.It provides data support in the data preparation stage of the comments generation model.Based on the constructed comments generation model and knowledge graph,this thesis designs and implements a long comments generation system for the automotive industry with web development technology.This system includes functions such as comments generation,comments publishing,knowledge graph expansion,and strategy recommendation.This system can not only serve as the comments writing assistant for operators to save labor costs,but also arouse the interest of potential consumers in buying cars through the generated comments.At the end of this thesis,various tests are carried out on the long comments generation system to verify the effectiveness of the system.