Abstract Despite many recent advances in automatic speech recognition (ASR), linguists and language communities engaged in language documentation projects continue to face the obstacle of the “transcription bottleneck”. Researchers in NLP typically do not distinguish between[...]