Search papers, labs, and topics across Lattice.
University of York
1
0
3
Benchmark models for Bangla hate speech detection fail spectacularly in real-world scenarios, with performance dropping by over 25% when faced with implicit expressions.