Research And Application Of Text Classification Method Based On Price Complaint Data

It is not only inefficient,but also requires a lot of time and energy to identify whether the whistleblower is a professional whistleblower.With the increasing application projects of natural language processing technology,the application of artificial intelligence technology constantly reflects the characteristics of diversity,especially in-depth learning,the application of natural language processing is deepening.In this technical background,based on the real pressure of the classification and processing of price complaint information in the actual work,this paper studies and analyzes the real complaint text data,in order to find a text classification algorithm suitable for the current actual work.Based on the classification models of convolutional neural network and cyclic neural network,this paper first introduces and analyzes the related technologies of text classification methods,studies the word segmentation methods,researches the extraction of high-frequency words and part of speech tagging for the original data,and then finds and convolutes the neural network through the research on the text classification application of the classic convolutional neural network algorithm The most suitable feature extraction model and convolution neural core parameters of the network algorithm,and choose two feature extraction models combined with convolution neural network algorithm for experiments,can automatically identify the meta features of each kind of data information,which is helpful for data classification.However,using a single model,the definition of the extracted data features is not clear enough,and the calculation time is too long.Therefore,the convolution neural network and the cyclic neural network are combined into a hybrid model to study,absorb the advantages of the two models,realize fast learning and accurately extract features.In order to optimize the weight of feature vector between the two models,attention model idea is introduced to improve the model and algorithm,more attention is paid to the extraction process of text classification features,and the invalid feature information is excluded,so as to improve the quality of features.On this basis,according to the semantic analysis,this paper puts forward the shortcomings and defects of different expressions of words and words in text classification,designs a two-channel word combination short text classification model based on convolution neural network and cyclic neural network,effectively excavates the key words or keywords in the text,and further improves the accuracy of text classification.Through the design and experiment of the classification method,the results show that the hybrid model based on convolutional neural network and cyclic neural network is the best for complaint text classification,with an accuracy of about 94%.Finally,the design and implementation of the complaint information classification system,the implementation effect can be seen that the system can directly display the effect of the classification model studied in this subject to the system users,which is convenient for the system users to quickly realize the text classification management through the functional modules of the operating system.