BART model fine-tuned to aggregate crowd-sourced transcriptions. Repository: GitHub