A multi-level information extraction model for end-to-end aspect-based sentiment analysis
Jiale Wang, Wensheng Sun · International Conference on Electronic Information Technology (EIT 2022) · 2022
Aspect-based sentiment analysis mainly processes natural language and generates its aspect term and corresponding sentiment. Previous studies focused on individual subtasks, did not make full use of large-scale corpus and dig out the semantic information of sentences, and did not consider the different contributions of different words in aspect-based sentiment analysis. In this paper, a unified sequence labeling BIO model is used, and the two subtasks of aspect word extraction and sentiment analysis are fused. By sharing a unified BERT model, and using the auxiliary training of domain-related documents, the attention is focused on aspect words and emotional words. Through experiments, a better effect was finally achieved.