Aspect and Opinion Term Extraction Using Graph Attention Network

TL;DR

The paper tackles Aspect Based Sentiment Analysis by reframing aspect and opinion term extraction as token-level NER and proposes Relational Graph Attention Networks that operate over dependency-structured graphs. It introduces an Aspect Oriented Dependency Tree to align the graph with aspect targets and combines RGAT with BERT token embeddings and various tagging heads (Linear, BiLSTM, Transformer) plus a CRF layer. The approach achieves state-of-the-art or near-state-of-the-art results across SemEval ABSA datasets, demonstrating strong gains in both aspect and opinion term extraction, including scenarios with multiple aspects. The work highlights the importance of syntactic structure in ABSA and offers a versatile framework that performs well across multiple domains, though it relies on language-specific dependency parsing and explicit alignment between aspects and opinions. Future directions include end-to-end matching of aspect-opinion pairs and exploring decoder architectures like pointer networks.

Abstract

In this work we investigate the capability of Graph Attention Network for extracting aspect and opinion terms. Aspect and opinion term extraction is posed as a token-level classification task akin to named entity recognition. We use the dependency tree of the input query as additional feature in a Graph Attention Network along with the token and part-of-speech features. We show that the dependency structure is a powerful feature that in the presence of a CRF layer substantially improves the performance and generates the best result on the commonly used datasets from SemEval 2014, 2015 and 2016. We experiment with additional layers like BiLSTM and Transformer in addition to the CRF layer. We also show that our approach works well in the presence of multiple aspects or sentiments in the same query and it is not necessary to modify the dependency tree based on a single aspect as was the original application for sentiment classification.