Aspect-level sentiment analysis model based on syntactic information and part of speech encoding
Yiqiu Fang, Peng Guo, Junwei Ge · 2022
Aspect-level sentiment analysis aims to predict the sentiment polarity of various aspect words in a sentence. Recent studies have pointed out that the sentiment polarity of aspect words often depends on the local context that is highly correlated with aspect words. However, there are still some problems in the existing work: a) syntactic information is not fully utilized; b) the method of weight assignment to local context is not scientific enough; c) the importance of parts of speech is ignored. Aiming at the existing problems, an aspect-level sentiment analysis model based on syntactic information and part-of-speech encoding is proposed. The model innovatively proposes a pre-training module that integrates syntactic information, introduces a part-of-speech encoding component, and designs a new dynamic weight assignment method. The proposed model is experimented on three publicly available datasets, and the experimental results show that the model achieves better results on a macro scale than existing models.