Training Entire-Space Models for Target-oriented Opinion Words Extraction

TL;DR

This paper addresses the sample selection bias in target-oriented opinion words extraction by training models on the entire space of instances, including those without opinion words, leading to improved performance and more accurate evaluation.

Contribution

It introduces a method for training TOWE models on the entire instance space, including cases without opinion words, and demonstrates the benefits over previous approaches that used only biased samples.

Findings

Training on entire space improves model performance.

Evaluating only on biased samples overestimates performance.

Including all instance types enhances generalization.

Abstract

Target-oriented opinion words extraction (TOWE) is a subtask of aspect-based sentiment analysis (ABSA). Given a sentence and an aspect term occurring in the sentence, TOWE extracts the corresponding opinion words for the aspect term. TOWE has two types of instance. In the first type, aspect terms are associated with at least one opinion word, while in the second type, aspect terms do not have corresponding opinion words. However, previous researches trained and evaluated their models with only the first type of instance, resulting in a sample selection bias problem. Specifically, TOWE models were trained with only the first type of instance, while these models would be utilized to make inference on the entire space with both the first type of instance and the second type of instance. Thus, the generalization performance will be hurt. Moreover, the performance of these models on the first type of instance cannot reflect their performance on entire space. To validate the sample selection bias problem, four popular TOWE datasets containing only aspect terms associated with at least one opinion word are extended and additionally include aspect terms without corresponding opinion words. Experimental results on these datasets show that training TOWE models on entire space will significantly improve model performance and evaluating TOWE models only on the first type of instance will overestimate model performance.