Research On Short Text Classification Technology Based On Deep Learning

Short text classification is the process of letting a computer distinguish a given text as one or several of several text categories determined in advance under a given classification system.Short text classification plays an important role in redundancy filtering,intelligent retrieval,indexing,text filtering and so on.It is convenient for users to solve problems quickly and efficiently.Because of the growing demand for short text classification,how to improve the accuracy of short text classification more efficiently has become a major challenge currently facing.With the emergence of the BERT model in recent years,many problems in the NLP task that cannot be solved by other models have been solved,which is bound to play a positive role in the development of NLP.The BERT model is extremely flexible,can deal with a variety of corpora,and provide huge results for the corpus.The BERT model is very popular,not only in the NLP field,but also by researchers in other industries.The BERT model has become recognized as the most influential pre-training model.This paper comprehensively analyzes the current difficulties and hot issues in short text classification research through the research on short text classification related technologies and other knowledge,and deep-level analysis improves the analysis of classification accuracy.The BERT model is a machine learning model that has become popular in recent years and is currently in the leading ranks.It can perform multiple operations at the same time,such as unsupervised text classification,supervised text classification,etc.It is naturally feasible to apply the BERT model to natural language processing.The research content of this paper mainly includes the following aspects:1)For the current short text feature extraction is difficult,short text key information is less,model training time is long,For problems such as poor generalization ability of the training model,this paper proposes a short text classification technology based on the BERT model.This model can systematically complete short text classification tasks,and directly output the processed text information to the BERT model.Short text can be classified without other operations,which is more convenient and quick to implement short text classification.By comparing the performance of other classification models on the same data set,this paper confirms that the BERT model has improved the short text classification research.2)In view of the short text classification deficiencies of the BERT model,this paper analyzes the characteristics of the support vector machine in the short text classification task,fuses the BERT model and the support vector machine classifier,and builds the S-BERT model to further compare The accuracy of short text classification is improved,so as to solve the short text classification task of the BERT model.In this article,on the public data set,the S-BERT model and other classification models(including the BERT model)are compared experimentally on the same data set,and finally it is concluded that the improvement of the BERT model can indeed improve the classification accuracy of the original BERT model.