Hierarchical Interactive Network for joint aspect extraction and sentiment classification

Chen, Wei; Lin, Peiqin; Zhang, Wanqi; Du, Jinglong; He, Zhongshi

doi:10.1016/j.knosys.2022.109825

Cited by 9 publications

(16 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this work, we aim to tackle the task of jointly predicting aspects and their corresponding sentiments. As shown in prior studies (Chen et al 2022b;Liang et al 2023;Liu et al 2023), AP and SP are highly related when applied to the same sentence, and thus a MTL solution can potentially benefit for both sub-task. It is important to note that our approach is not restricted to these two sub-tasks alone, but can be generally extended to other MTL problems as well.…”

Section: Problem Formulationmentioning

confidence: 76%

“…Recent efforts aim to achieve the two sub-tasks cohesively, predicting aspects and corresponding sentiments jointly or interactively through a unified methodology (Lin and Yang 2020;Lv et al 2021;Chen et al 2022b;Zhang et al 2022c;Liang et al 2023;Cui et al 2023). These studies have revealed a strong mutual implication between these two sub-tasks, and have devised some interaction strategies to facilitate the knowledge exchange.…”

Section: Sentencesmentioning

confidence: 99%

“…Early studies treat them as two separate tasks and use some prevalent algorithms such as Conditional Random Fields (CRF) (Li and Lu 2017), Long Short-Term Memory (LSTM) (Chen et al 2020) and Bidirectional Encoder Representations from Transformers (BERT) (Yang et al 2020) to complete the task. Meanwhile, some researchers have explored end-to-end solutions for aspects and sentiments extraction using the multi-task learning framework (Luo et al 2019;Lin and Yang 2020;Chen et al 2022b;Liang et al 2023). They emphasize the close relationship between aspect and sentiment and propose various interaction approaches, such as shared-private feature interaction.…”

Section: Related Workmentioning

confidence: 99%

“…They emphasize the close relationship between aspect and sentiment and propose various interaction approaches, such as shared-private feature interaction. (Lin and Yang 2020), feature combination interaction (Lv et al 2021), and task-regularization interaction (Chen et al 2022b). Although successful, most of these methods overlook the varying degree of feature correlations in different sentences, making it challenging to transfer knowledge effectively across tasks.…”

Section: Related Workmentioning

confidence: 99%

“…Following (Chen et al 2022b;Lv et al 2023), we adopt pretrained BERT language model (Devlin et al 2019) as the underlying encoder to generate contextualized clause representations. Given a sentence x = {w 1 , w 2 , .…”

Section: Base Structurementioning

confidence: 99%

See 4 more Smart Citations

Modeling Adaptive Inter-Task Feature Interactions via Sentiment-Aware Contrastive Learning for Joint Aspect-Sentiment Prediction

Chen,

Liu,

Zhang

et al. 2024

AAAI

View full text Add to dashboard Cite

Aspect prediction (AP) and sentiment prediction (SP) are representative applications in fine-grained sentiment anal- ysis. They can be considered as sequential tasks, where AP identifies mentioned aspects in a sentence, and SP infers fine-grained sentiments for these aspects. Recent models perform the aspect-sentiment prediction in a joint man-ner, but heavily rely on the feature interactions of aspect and sentiment. One drawback is that they ignore correlation strength varies between aspect features and sentiment fea- tures across different sentences, and employ a fixed feature interaction strategy may limit effective knowledge transfer across tasks. To tackle this issue, in this paper, we propose an Adaptive Inter-task Feature Interaction framework, AIFI, for joint aspect-sentiment prediction. Specifically, we introduce a novel contrast-based alignment method based on contrastive learning. Our approach considers the AP-specific and SP-specific representations of a given sentence as a positive pair, while representation of another random sentence serves as a negative example. Moreover, we propose an inter-task feature correlation network to predict the contrast strength, which is determined by the temperature coefficient in the InfoNCE loss. This dynamic correlation adjustment enhances model’s ability to capture proper feature interactions more efficiently. Experimental results on three datasets validate the effectiveness of our approach.

show abstract

Section: Problem Formulationmentioning

confidence: 76%