Datasets:

---

msmaje

/

hatadataset-cleaned-accordingly

like 0

HATA-cleaned corpus: 447,973 records | 6 targeted passes (normalise, JSON extract, URL remove, Arabic remove, boilerplate remove, exact dedup) | spam/adult/short text RETAINED for HATA | derived from msmaje/expanded_hausa_corpus