An Uzbek Medical-Domain Dataset for Aspect-Based Sentiment Analysis

UzMedSentiment is a manually annotated Uzbek medical-domain dataset designed for sentiment classification, aspect-based sentiment analysis, and auxiliary linguistic signal detection.

This dataset contains 4,791 annotated text instances collected from various online sources such as Telegram, Instagram, forums, and web comments. Each instance is provided in a structured tabular format and includes the following fields: text content, aspect category, sentiment label, polarity score, and source metadata.

In addition, the dataset provides rich linguistic and clinical annotations, including adverse drug reaction (ADR) flags, severity levels, negation, speculation, sarcasm, and cue-span information.

The dataset is intended to support research in Uzbek natural language processing (NLP), clinical text mining, opinion mining, and low-resource language technologies. It can be used for developing and evaluating machine learning and deep learning models for medical sentiment analysis and fine-grained linguistic signal detection.

Data format: The dataset is provided in both XLSX and TSV formats. Each row corresponds to a single annotated instance, and the TSV version is recommended for machine learning and NLP tasks.