Quality estimation of hybrid transcription of audio

摘要: Hybrid transcription of audio relies on having one or more layers transcribers who review transcriptions generated by automatic speech recognition (ASR) systems in order to correct errors that are found the transcriptions. When it comes determining how much human reviewing is needed, such as many use, there a cost/benefit tradeoff consider. Some embodiments described herein utilize machine learning-based approach for estimating quality hybrid audio. In embodiment, computer generates segment using an ASR system, which subsequently reviewed transcriber. The then calculates, based properties transcriber, value indicative expected accuracy transcription. may suggest second transcriber if below threshold.