Update README.md

README.md

@@ -1,3 +1,8 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+---
+AbuseBERT is a BERT-based model fine-tuned for abusive language detection.  
+It leverages ten diverse datasets harmonized under a unified taxonomy to improve cross-dataset generalization.  
+The integrated approach mitigates sampling and lexical biases, achieving superior F1 scores on held-out benchmarks compared to models trained on individual datasets.  
+
+AbuseBERT is designed to be **robust and generalizable across platforms, domains, and contexts**, making it well-suited for research, social media moderation studies, and real-world abusive language detection tasks.