hausasent-hausa-sentiment

HausaSent Hausa sentiment

3-class sentiment model (negative, neutral, positive) for Hausa and Hausa–English code-mixed text.

Fine-tuned from Davlan/afro-xlmr-mini in the HausaSent teaching/portfolio project.

Training data

• Extra synthetic Hausa / code-mixed templates (market, clinic, school, civic, social). Synthetic text is for coverage, not for published claims.
• Human-labeled NaijaSenti Hausa tweets (CC BY-NC-SA 4.0).

Use

from transformers import pipeline

clf = pipeline("text-classification", model="Abduull6771/hausasent-hausa-sentiment")
print(clf("Na ji daɗi da sabis ɗin."))

Do not deploy on clinic, school, or civic messages without a human in the loop. Evaluate on your own labeled Hausa before any performance claim.

Citation

If you use the NaijaSenti portion, cite Muhammad et al. (2022), NaijaSenti: A Nigerian Twitter Sentiment Corpus for Multilingual Sentiment Analysis.