in22-gen

IN22-Gen IN22 is a newly created comprehensive benchmark for evaluating machine translation performance in multi-domain, n-way parallel contexts across 22 Indic languages. IN22-Gen is a general-purpose multi-domain evaluation subset of IN22. It has been created from two sources: Wikipedia and Web Sources offering diverse content spanning news, entertainment, culture, legal, and India-centric topics. The evaluation subset consists of 1024 sentences translated across 22 Indic… See the full description on the dataset page:

Source: Hugging Face Hub (ai4bharat/IN22-Gen). Metadata imported from the dataset’s Hub tags.