Dynamic Order Template Prediction for Generative Aspect-Based Sentiment Analysis

TL;DR

This paper introduces a Dynamic Order Template (DOT) approach for aspect-based sentiment analysis that dynamically generates views for each instance, improving accuracy and efficiency over static templates and multi-view prompting methods.

Contribution

The paper proposes a novel dynamic template generation method for ABSA that adapts to each instance, addressing limitations of static templates and multi-view prompting.

Findings

Improves F1-scores on ASQP and ACOS datasets.

Reduces inference time significantly.

Enhances dependency capturing between sentiment tuple elements.

Abstract

Aspect-based sentiment analysis (ABSA) assesses sentiments towards specific aspects within texts, resulting in detailed sentiment tuples. Previous ABSA models often use static templates to predict all of the elements in the tuples, and these models often fail to accurately capture dependencies between elements. Multi-view prompting method improves the performance of ABSA by predicting tuples with various templates and then ensembling the results. However, this method suffers from inefficiencies and out-of-distribution errors. In this paper, we propose a Dynamic Order Template (DOT) method for ABSA, which dynamically generates necessary views for each instance based on instance-level entropy. Ensuring the diverse and relevant view generation, our proposed method improves F1-scores on ASQP and ACOS datasets while significantly reducing inference time.