Sentiment analysis of agricultural product ecommerce review data based on deep learning

The relevant comment texts of agricultural products sold on the e-commerce platform contain various and complex emotions of consumers, and reflect customers' consumer experience and views on those products. Compared with the features and advantages of English text segmentation structure, the analysis of Chinese sentiment polarity in a complex context has always been a problem in deep learning. This article explores the issue of sentiment polarity analysis of short texts of Chinese reviews of agricultural products by constructing Text-CNN, Bi-LSTM and BERT to fine-tune three deep learning models. The research object adopted is the review data of agricultural products on the e-commerce platform obtained based on crawler technology. The experimental results show that, compared with the Bi-LSTM and BERT fine-tuning models, the Text-CNN model's accuracy, precision, recall, and F1 value are all about 3 to 8 percentage points higher. The classification effect gradient gap between the three models is obvious. The accuracy of the Text-CNN convolutional neural network reached 99.92%. The results confirmed that Text-CNN has high sentiment analysis accuracy and good sentiment polarity classification effect in processing agricultural short text comment data, as an agricultural product e-commerce The emotional classification model of sales review data can provide an important reference for manufacturers to accurately distinguish user feedback and enhance market competitiveness.