Open-Domain Targeted Sentiment Analysis via Span-Based Extraction and Classification

TL;DR

This paper introduces a span-based extract-then-classify framework for open-domain targeted sentiment analysis, addressing issues of search space and sentiment inconsistency in previous sequence tagging methods, and demonstrates superior performance on benchmark datasets.

Contribution

Proposes a novel span-based framework with three model variants for targeted sentiment analysis, outperforming traditional sequence tagging approaches.

Findings

The span-based approach outperforms sequence tagging baselines.

The pipeline model achieves the best performance among the proposed models.

Experiments on three datasets validate the effectiveness of the approach.

Abstract

Open-domain targeted sentiment analysis aims to detect opinion targets along with their sentiment polarities from a sentence. Prior work typically formulates this task as a sequence tagging problem. However, such formulation suffers from problems such as huge search space and sentiment inconsistency. To address these problems, we propose a span-based extract-then-classify framework, where multiple opinion targets are directly extracted from the sentence under the supervision of target span boundaries, and corresponding polarities are then classified using their span representations. We further investigate three approaches under this framework, namely the pipeline, joint, and collapsed models. Experiments on three benchmark datasets show that our approach consistently outperforms the sequence tagging baseline. Moreover, we find that the pipeline model achieves the best performance compared with the other two models.