The Somali dataset doesn't have correct data
#4
by
muhaksim
- opened
I checked the subset dataset which is labeled "SO". I think it is referring to Somali. There is not any useful data there that makes sense.
Am I missing something here?
Yes, SO refers to Somali. We noticed this issue too. We inherited the data for this language from OSCAR so it's actually small and can be improved. This is our ongoing work and we expect to improve our pipeline to include larger and better data here, especially for low-resource languages, as we move forward.