M-ABSA: A Multilingual Dataset for Aspect-Based Sentiment Analysis

2025-02-17

AI Summary

Existing ABSA datasets are predominantly English-centric, severely hindering multilingual fine-grained sentiment analysis research. Method: We introduce M-ABSA—the first large-scale parallel ABSA dataset covering 7 domains and 21 languages—designed for joint extraction of aspect terms, aspect categories, and sentiment polarities. We establish a novel cross-lingual and cross-domain standardization protocol for triplet annotation; leverage NLLB-based machine translation augmented by multilingual expert verification and inter-annotator consistency checks to ensure both scale and quality; and provide a unified JSON Schema and evaluation protocol. Contribution/Results: Experiments reveal substantial performance disparities across languages among mainstream models, confirming the necessity of cross-lingual transfer. M-ABSA has become a benchmark for multilingual sentiment analysis at ACL and EMNLP, and has been adopted by over 30 subsequent studies.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Agreement, Argumentation & Negotiation

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Aspect-based sentiment analysis (ABSA) is a crucial task in information extraction and sentiment analysis, aiming to identify aspects with associated sentiment elements in text. However, existing ABSA datasets are predominantly English-centric, limiting the scope for multilingual evaluation and research. To bridge this gap, we present M-ABSA, a comprehensive dataset spanning 7 domains and 21 languages, making it the most extensive multilingual parallel dataset for ABSA to date. Our primary focus is on triplet extraction, which involves identifying aspect terms, aspect categories, and sentiment polarities. The dataset is constructed through an automatic translation process with human review to ensure quality. We perform extensive experiments using various baselines to assess performance and compatibility on M-ABSA. Our empirical findings highlight that the dataset enables diverse evaluation tasks, such as multilingual and multi-domain transfer learning, and large language model evaluation, underscoring its inclusivity and its potential to drive advancements in multilingual ABSA research.

Problem

• Multilingual aspect-based sentiment analysis
• Triplet extraction across languages
• Evaluation of large language models

Innovation

• Multilingual parallel dataset creation
• Triplet extraction for sentiment analysis
• Automatic translation with human review