Bharat Parallel Corpus Collection (BPCC) is a comprehensive and publicly available parallel corpus that includes both existing and new data for all 22 scheduled Indic languages. It is comprised of two parts: BPCC-Mined and BPCC-Human, totaling approximately 230 million bitext pairs. BPCC-Mined contains about 228 million pairs, with nearly 126 million pairs newly added as a part of this work. On the other hand, BPCC-Human consists of 2.2 million gold… See the full description on the dataset page:
https://huggingface.co/datasets/ai4bharat/BPCC.
