Telegram News Sentiment and Topic Classification Dataset

Dataset of news messages scores collected from public Telegram channels. Each record includes the source channel, message ID, publication date (Unix timestamp), sentiment scores (positive, neutral, negative) obtained via a zero-shot classification model, and topic classification scores across eight thematic categories. This dataset supports the paper "From Noisy News Sentiment Scores to Interpretable Temporal Dynamics: A Bayesian State-Space Model".