iACOS: Advancing Implicit Sentiment Extraction with Informative and Adaptive Negative Examples

Xiancai Xu,Jia-Dong Zhang,Lei Xiong,Zhishang Liu

2024-06-22

Abstract:Aspect-based sentiment analysis (ABSA) have been extensively studied, but little light has been shed on the quadruple extraction consisting of four fundamental elements: aspects, categories, opinions and sentiments, especially with implicit aspects and opinions. In this paper, we propose a new method iACOS for extracting Implicit Aspects with Categories and Opinions with Sentiments. First, iACOS appends two implicit tokens at the end of a text to capture the context-aware representation of all tokens including implicit aspects and opinions. Second, iACOS develops a sequence labeling model over the context-aware token representation to co-extract explicit and implicit aspects and opinions. Third, iACOS devises a multi-label classifier with a specialized multi-head attention for discovering aspect-opinion pairs and predicting their categories and sentiments simultaneously. Fourth, iACOS leverages informative and adaptive negative examples to jointly train the multi-label classifier and the other two classifiers on categories and sentiments by multi-task learning. Finally, the experimental results show that iACOS significantly outperforms other quadruple extraction baselines according to the F1 score on two public benchmark datasets.

Computation and Language

What problem does this paper attempt to address?

### Problems the Paper Aims to Solve

This paper aims to address the issue of extracting implicit aspects and opinions in Aspect-based Sentiment Analysis (ABSA). Specifically:

1. **Extraction of Implicit Aspects and Opinions**:
- Existing research mostly focuses on the extraction of explicit aspects, neglecting the implicit aspects and opinions that do not directly appear in the text. This paper proposes a new method, iACOS, to simultaneously extract both explicit and implicit aspects and opinions.

2. **Joint Prediction of Quadruples**:
- A multi-label classifier is proposed to simultaneously predict the category and sentiment polarity of each aspect-opinion pair, thereby avoiding the error propagation problem in traditional pipeline methods.

3. **Informative and Adaptive Negative Sample Construction**:
- Informative and adaptive negative samples are used for multi-task training to improve the performance of quadruple extraction. These negative samples not only help the model distinguish subtle differences but are also dynamically generated based on the current model parameters, ensuring the model is continuously challenged.

Through the above methods, iACOS significantly outperforms other quadruple extraction techniques in experiments on two public benchmark datasets.