Research On Aspect-level Sentiment Classification Method Based On Deep Learning

Nowadays,with the rapid development of Internet technology,there is a large amount of text data on the Internet,all of which contain subjective emotions.In order to effectively mine the potential emotional information,sentiment analysis technology came into being.Although in most cases,affective analysis is very useful at both document and sentence levels,when a document or a sentence involves multiple emotional expressions,the first two levels of affective analysis will not be able to accurately extract the deep emotions in the text.Aspect-level affective analysis has been widely studied because of its more targeted mining of affective polarity and more accurate extraction of deep semantic features of text.Most aspect-level sentiment classification networks,including long-term and short-term memory(LSTM)networks,together with attention mechanism and memory module,have been widely used in aspect level emotion classification.Although it has achieved good results,it can not extract the global and local information of the context at the same time.It only models based on the semantic correlation between aspects and their corresponding context words,ignoring their syntactic dependence.Therefore,this paper proposes two syntactic dependency methods for aspect level emotion analysis based on deep learning.The research content mainly includes the following two parts:(1)This paper proposes the aspect-level sentiment classification by combining convolutional neural network(CNN)and proximity-weighted convolution network(PWCN),as well as a new method to calculate the proximity weight.To obtain contextualized word vectors,corpora has been trained by the model of bidirectional encoder representations from transformers(BERT),which can be taken as text features.The CNN is able to extract sequence features from the text and to take the sequence information from the text into account.In addition,the PWCN can consider the syntactic dependencies inside the sentences.The BERT model also has the ability to model complex features of words,such as their syntactic and semantic changes in a linguistic context.Experiments conducted on the Sem Eval 2014 benchmark demonstrate compared to the well-established ones,the proposed approach had bigger effectiveness.(2)An aspect level emotion classification method based on disconnected gated recurrent units(DGRU)and proximity weighted convolution(PWCN)is proposed.DGRU can capture both global sentence information and local emotion information in the text.Proximity weighted convolution(PWCN)can better capture syntactic feature information.Experiments conducted on the Sem Eval 2014 benchmark yielded relatively competitive results.