Investigating Typed Syntactic Dependencies for Targeted Sentiment Classification Using Graph Attention Neural Network
This work addresses a specific bottleneck in sentiment analysis for researchers and practitioners by improving accuracy in predicting sentiment on target mentions.
The paper tackles the problem of targeted sentiment classification by integrating typed syntactic dependency information into a graph attention neural network, resulting in a model that significantly outperforms state-of-the-art syntax-based approaches on standard benchmarks.
Targeted sentiment classification predicts the sentiment polarity on given target mentions in input texts. Dominant methods employ neural networks for encoding the input sentence and extracting relations between target mentions and their contexts. Recently, graph neural network has been investigated for integrating dependency syntax for the task, achieving the state-of-the-art results. However, existing methods do not consider dependency label information, which can be intuitively useful. To solve the problem, we investigate a novel relational graph attention network that integrates typed syntactic dependency information. Results on standard benchmarks show that our method can effectively leverage label information for improving targeted sentiment classification performances. Our final model significantly outperforms state-of-the-art syntax-based approaches.