Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Train Once for All: A Transitional Approach for Efficient Aspect Sentiment Triplet Extraction

Xinmeng Hou, Lingyue Fu, Chenhao Meng, Kounianhua Du, Wuqi Wang, Hai Hu

TL;DR

This work proposes the first transition-based model for AOPE and ASTE that performs aspect and opinion extraction jointly, which also better captures position-aware aspect-opinion relations and mitigates entity-level bias by integrating contrastive-augmented optimization.

Abstract

Aspect-Opinion Pair Extraction (AOPE) and Aspect Sentiment Triplet Extraction (ASTE) have drawn growing attention in NLP. However, most existing approaches extract aspects and opinions independently, optionally adding pairwise relations, often leading to error propagation and high time complexity. To address these challenges and being inspired by transition-based dependency parsing, we propose the first transition-based model for AOPE and ASTE that performs aspect and opinion extraction jointly, which also better captures position-aware aspect-opinion relations and mitigates entity-level bias. By integrating contrastive-augmented optimization, our model delivers more accurate action predictions and jointly optimizes separate subtasks in linear time. Extensive experiments on 4 commonly used ASTE/AOPE datasets show that, while performing worse when trained on a single dataset than some previous models, our model achieves the best performance on both ASTE and AOPE if trained on combined datasets, outperforming the strongest previous models in F1-measures (often by a large margin). We hypothesize that this is due to our model's ability to learn transition actions from multiple datasets and domains. Our code is available at https://anonymous.4open.science/r/trans_aste-8FCF.

Train Once for All: A Transitional Approach for Efficient Aspect Sentiment Triplet Extraction

TL;DR

This work proposes the first transition-based model for AOPE and ASTE that performs aspect and opinion extraction jointly, which also better captures position-aware aspect-opinion relations and mitigates entity-level bias by integrating contrastive-augmented optimization.

Abstract

Aspect-Opinion Pair Extraction (AOPE) and Aspect Sentiment Triplet Extraction (ASTE) have drawn growing attention in NLP. However, most existing approaches extract aspects and opinions independently, optionally adding pairwise relations, often leading to error propagation and high time complexity. To address these challenges and being inspired by transition-based dependency parsing, we propose the first transition-based model for AOPE and ASTE that performs aspect and opinion extraction jointly, which also better captures position-aware aspect-opinion relations and mitigates entity-level bias. By integrating contrastive-augmented optimization, our model delivers more accurate action predictions and jointly optimizes separate subtasks in linear time. Extensive experiments on 4 commonly used ASTE/AOPE datasets show that, while performing worse when trained on a single dataset than some previous models, our model achieves the best performance on both ASTE and AOPE if trained on combined datasets, outperforming the strongest previous models in F1-measures (often by a large margin). We hypothesize that this is due to our model's ability to learn transition actions from multiple datasets and domains. Our code is available at https://anonymous.4open.science/r/trans_aste-8FCF.

Paper Structure

This paper contains 31 sections, 11 equations, 3 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Previous methods on ASTE
  4. Transition-based Methods in NLP
  5. Contrastive-based Optimization
  6. The Trans-AOPE/ASTE Model
  7. Transitional Operations and State Change
  8. Default Actions:
  9. Primary Actions:
  10. Secondary Actions:
  11. Trans-AOPE State Representation
  12. Transition Implementation with Neural Model
  13. Token representations
  14. State Initialization
  15. Step-by-Step Parser State Representation.
  16. ...and 16 more sections

Figures (3)

  • Figure 1: Demonstration of the processing steps in both classic and transitional methods for extracting aspect-opinion pairs and tagging sentiment polarity. Importantly, our proposed transitional method predicts transition actions, and performs pair extraction after the aspect–opinion relationship has been established, allowing the model to capture contextual relationships more effectively. AE stands for Aspect Extraction, OE for Opinion Extraction. POS and NEG represent positive and negative sentiments respectively.
  • Figure 2: The complete process of the transition-based model is illustrated. Purple highlights represent the transition-based pair extraction, while orange indicates the final step of sentiment tagging.
  • Figure 3: F1 score as a function of training epochs in the combined-train condition for the AOPE task on the 14lap test set, with various loss weight configurations. w1=base loss; w2=contrastive loss.