- https://doi.org/10.1145/3766671.3766765
Research on Text Sentiment Analysis Model Based on Graph Attention Network
- Jun 13, 2025
- Mingchu Yu +1 more
In this paper, we introduce text sentiment analysis model GAT-SA (Graph Attention Network for Sentiment Analysis) to better recognize subtle emotional expressions. The conventional way for sentiment analysis lacks of deep semantic relationship between words, which is the key point to analyzing some fine-grained emotions. Therefore, we first make word text to a form of graph, the node presents words and the edge presents syntactic dependencies. With graph attention mechanism, GAT dynamically adjust the weight between neighbors so that it can better capture sentiment. With the multi-head attention mechanism, GAT-SA can better capture long range dependency and context. Experimental results indicate that GAT-SA performs better on sentiment analysis dataset IMDB, Twitter, with higher accuracy and F1 scores compared with traditional methods.
- # Graph Attention Mechanism
- # Multi-head Attention Mechanism
- # Sentiment Analysis
- # Graph Attention
- # Range Dependency
- # Attention Mechanism
- # Subtle Expressions
- # Deep Relationship
- # Higher Accuracy
- # Accuracy Scores
