DimABSA: Building Multilingual and Multidomain Datasets for Dimensional Aspect-Based Sentiment Analysis

TL;DR

DimABSA introduces the first multilingual, dimensional ABSA resource annotated with valence–arousal (VA) scores alongside traditional ABSA elements, covering 76,958 aspect instances across 42,590 sentences in six languages and four domains. It defines three subtasks (DimASR, DimASTE, DimASQP) to bridge categorical ABSA to continuous sentiment, and proposes the continuous F1 (cF1) metric to jointly evaluate categorical extraction/classification and VA regression. Through extensive prompting and fine-tuning of large language models, the study demonstrates the dataset's challenge and the value of supervised adaptation, while highlighting language-resource disparities and the need for robust cross-lingual dimensional ABSA methods. DimABSA thus provides a foundation for fine-grained, multilingual sentiment analysis with practical applicability to multilingual NLP tasks and cross-domain analyses.

Abstract

Aspect-Based Sentiment Analysis (ABSA) focuses on extracting sentiment at a fine-grained aspect level and has been widely applied across real-world domains. However, existing ABSA research relies on coarse-grained categorical labels (e.g., positive, negative), which limits its ability to capture nuanced affective states. To address this limitation, we adopt a dimensional approach that represents sentiment with continuous valence-arousal (VA) scores, enabling fine-grained analysis at both the aspect and sentiment levels. To this end, we introduce DimABSA, the first multilingual, dimensional ABSA resource annotated with both traditional ABSA elements (aspect terms, aspect categories, and opinion terms) and newly introduced VA scores. This resource contains 76,958 aspect instances across 42,590 sentences, spanning six languages and four domains. We further introduce three subtasks that combine VA scores with different ABSA elements, providing a bridge from traditional ABSA to dimensional ABSA. Given that these subtasks involve both categorical and continuous outputs, we propose a new unified metric, continuous F1 (cF1), which incorporates VA prediction error into standard F1. We provide a comprehensive benchmark using both prompted and fine-tuned large language models across all subtasks. Our results show that DimABSA is a challenging benchmark and provides a foundation for advancing multilingual dimensional ABSA.

Figures (10)

• Figure 1: Valence–Arousal (VA) space. Illustrated examples of aspect instances across multiple languages and domains. Selected instances are visualized as triplets (aspect, opinion, paired VA score). Blue dots show a U-shaped distribution of VA scores.
• Figure 2: Aspect-Category distributions for the restaurant and laptop domains aggregated across languages.
• Figure 3: Joint Valence-Arousal (VA) distributions across various languages in the DimABSA dataset.
• Figure 4: Few-shot Performance of GPT-5 mini. Results are summarized across all tasks and datasets. See Appendix F for corresponding numerical scores.
• Figure 5: Gold and predicted VA distributions for DimASR. Results represent the English–Restaurant test set; see Appendix \ref{['Few-shotVAscatter']} for other datasets.