Sentiment Analysis by Double Classification of Takeaway Platform Reviews Based on Deep Learning LSTM Models

Abstract

Sentiment analysis has a wide range of applications in the fields of opinion analysis, sentiment dialog, and product reviews. However, the sentiment information expressed in texts under different topics varies greatly; for example, a model that performs well on a movie review set has poor model classification on a social platform review set due to inconsistent recognition of antiphonal phrases, different expression of emoji sentiment, and missing contextual information. In this paper, the authors focus on tens of thousands of latest reviews of Chinese takeout platforms Meituan and Elema, and use the LSTM model in deep learning to double classify the data (positive and negative). This paper analyzes the performance of LSTM models in the field of sentiment analysis of takeout reviews and concludes that domain-specific text sentiment analysis requires specific analysis.