关键词: 情感分析, 情感词典, 机器学习, 深度学习, 注意力机制, 预训练

Abstract:

Text sentiment analysis is an important branch of natural language processing, which is widely used in public opinion analysis and content recommendation. It is also a hot topic in recent years. According to different methods used, it is divided into sentiment analysis based on emotional dictionary, sentiment analysis based on traditional machine learning, and sentiment analysis based on deep learning. Through comparing these three methods, the research results are analyzed, and the paper summarizes the advantages and disadvantages of different methods, introduces the related data sets and evaluation index, and application scenario, analysis of emotional subtasks is simple summarized. The future research trend and application field of sentiment analysis problem are found. Certain help and guidance are provided for the researchers in the related areas.

Key words: sentiment analysis, emotional dictionary, machine learning, deep learning, attention mechanism, pre-training