New Sinhala Dataset Enables Fine-Grained Market Sentiment Analysis
The SalAngaBhava dataset allows NLP models to move beyond general sentiment to specific product aspects in a low-resource language.
The SalAngaBhava dataset provides a structured way to perform aspect-based sentiment analysis (ABSA) for the Sinhala language. By manually labeling product reviews with specific aspect terms and associated sentiments—positive, negative, or neutral—the research enables models to identify exactly which features of a product are driving consumer opinion.
Most sentiment tools for low-resource languages only predict sentiment at the sentence level. This work introduces a more granular approach by pairing specific aspects with sentiments across several domains, filling a critical gap for an Indo-Aryan language used primarily in Sri Lanka.
https://arxiv.org/abs/2607.05259
The dataset uses user-generated reviews and comments annotated under strict guidelines to ensure balance and quality. This creates a reliable benchmark for developing NLP tools that can parse complex, domain-specific feedback in a market previously underserved by high-resolution sentiment data.
Expanding these capabilities to low-resource languages allows for more precise consumer intelligence in regional markets.