Logo Lanfrica

ssa-mte

Domain:

natural language processing

Record type:

dataset
Creator:
McG
Host:
We introduce SSA-MTE, a large-scale human-annotated MT evaluation (MTE) dataset. Link to the paper: SSA-COMET: Do LLMs Outperform Learned Metrics in Evaluating MT for Under-Resourced African Languages? SSA-COMET: Do LLMs Outperfo… The quality of data for each language is discussed in the paper. The data for 4 languages did not pass the quality check, and we are working on further filtering of them to enhance the final quality. (Swahili, Sesotho, Igbo, Zulu). Coming soon.