Automatic Speech Recognition
Automatic Speech Recognition — The automated conversion of spoken audio into text, typically combining acoustic modeling, language constraints, decoding, and segmentation logic.
Boundaries
ASR transcribes speech; speaker diarization attributes time regions to speakers. The two tasks can be combined but are not interchangeable.
An ASR system may output words, timestamps, punctuation, and speaker labels, but these are not necessarily produced by the same mechanism or with equal reliability. Word accuracy, temporal alignment, and speaker attribution require separate references and evaluations.
Related Concepts
Related article: Automatic Speech Recognition