Sentiment Analysis Method for Agricultural Product Review Based on Corpus Characteristics and Deep Learning Model

Abstract

Although deep learning models are widely used in text sentiment analysis, it is a challenging task to extract richer semantic features to improve model performance in corpora with weak label characteristics. This study crawls the agricultural product review of Jingdong e-commerce as a corpus, and proposes a deep learning method based on the characteristics of the corpus for sentiment analysis. The method first uses frequent item mining to construct a sentiment dictionary, and converts weakly labeled data into high-quality corpus through sentiment value calculation. Secondly, Convolutional Neural Network (CNN) and Bidirectional Long Short Term Memory (BiLSTM) are combined in the sentiment analysis model, and the word vectors trained by Glove and Word2vec are imported into the multi-channel neural network, so that the model can learn local and global semantic features in parallel, and embed the attention mechanism in the channel. The experimental results show that the performance of the model considering the characteristics of the corpus is significantly improved, and the MAtt-CNN-BiLSTM model constructed in this paper has the best performance in the experiments under the three datasets.