HAMOS: A Fine-Grained Hotel Aspect-Based Multimodal Opinion and Sentiment Dataset
Published: 17 August 2026| Version 3 | DOI: 10.17632/x46f4y4fpr.3
Contributor:
Description
HAMOS is a multimodal, multilingual dataset for fine-grained aspect-based sentiment analysis of hotel reviews. Each review is paired with its images and original captions, and every opinion is annotated as a sentiment quadruple — (aspect term, aspect category, opinion term, sentiment polarity) — under a taxonomy of six aspect categories, 31 sub-aspects, and three polarities, with emojis preserved and explicit/implicit expressions distinguished. It contains 3,399 hotels, 8,796 reviews, 12,601 text segments, 9,219 images, and 23,995 quadruples across English, Vietnamese, and other languages, collected from Booking.com and split into hotel-disjoint train/validation/test sets. Data are provided in JSON, JSONL, and CSV formats.
Files
Institutions
- Ho Chi Minh City University of ScienceHo Chi Minh, Ho Chi Minh City
Categories
Natural Language Processing, Machine Learning, Hospitality Management, Sentiment Analysis, Multimodal Learning