GSent

# GSent-98K: Gujarati News Headlines Sentiment Dataset ## Dataset DescriptionGSent-98K is a large-scale dataset of Gujarati news headlines annotated for sentiment: positive, negative, and neutral. - **Full dataset**: 88,223 semi-manual annotations - **Manual subset**: 10,329 fully manually annotated headlines for quality validation ## Columns- `Headline`: Gujarati news headline - `Label`: Sentiment (-1 = Negative, 0 = Neutral, 1 = Positive) ## Annotation Guidelines- **Positive**: Headlines expressing positive sentiment or praise - **Negative**: Headlines expressing criticism, concern, or negative sentiment - **Neutral**: Informative or factual headlines with no clear sentiment ## Versioning- v1.0 semi-manual: Full semi-automated annotations - v1.0 manual: Gold-standard manually annotated subset ## Ethical Considerations- Dataset is for research purposes only - Avoid harmful or discriminatory use ## Examples| Headline | Label ||----------|-------|| “ગુજરાતમાં શાળાઓ ફરી ખોલાય છે” | 1 || “મહામારીને કારણે બજારો બંધ છે” | -1 || “રાજ્ય સરકારનું નવું બજેટ જાહેર” | 0 | ## How to Use- Load in Python using pandas:```pythonimport pandas as pdfull = pd.read_csv('data/GSent88K_full.csv', encoding='utf-8-sig')manual = pd.read_csv('data/GSent88K_manual.csv', encoding='utf-8-sig')