Corpus of Indonesian Online Gambling Promotion Texts (CIOGPT)

Dataset Description This dataset contains 11,673 comments scraped from 24 diverse YouTube videos covering topics like entertainment, technology, and travel. The primary goal is to provide data for developing and evaluating models that can detect gambling-related content. Dataset Structure The comments are divided into two categories, with gambling comments representing a minority class (approximately 10%). The dataset is split into three subsets for model training and evaluation: Training Set: 8,171 comments (70%) Test Set: 2,335 comments (20%) Holdout Set: 1,167 comments (10%) Content & Annotation The dataset includes both gambling and non-gambling comments. Gambling Comments: Typically contain promotional text, winning claims, or withdrawal testimonials. Non-Gambling Comments: Include general discussions and casual observations. All data points were carefully labeled by multiple annotators to ensure high reliability, with their agreement measured using the Cohen's kappa coefficient.