liveBook · Manning

Word error rate

Word error rate is the ratio of word-level substitutions, deletions, and insertions required to transform a recognized transcription into a reference transcription.

How 2 Manning books explain it

Sutskever's List

Word error rate is the proportion of word-level recognition errors, typically computed from substitutions, deletions, and insertions relative to a reference transcript.

How it is used
Evaluates speech-recognition systems by comparing word-level errors across datasets and against human performance.
How deep it goes
Defines the error proportion and compares reported results, but does not derive alignment procedures or calculation details.
Examples
Uses WSJ, LibriSpeech, VoxForge, and CHiME results to compare DS2 with human listeners.
What you can do after
Evaluate speech-recognition quality across datasets and identify when environmental conditions widen the human–system gap.

Read how Sutskever's List explains word error rate →

Natural Language Processing in Action, Second Edition

Word error rate is the proportion of word-level transcription errors relative to the reference transcription.

How it is used
Evaluates and compares automatic speech-to-text accuracy when selecting a system for a voice assistant.
How deep it goes
Defines word-level transcription comparison and reports benchmark results, but does not detail the calculation procedure.
Examples
Uses AI, phone, meeting, video, and finance benchmark datasets to compare Kaldi, wav2vec 2.0, and Whisper.
What you can do after
Evaluate STT systems using benchmark WERs and compare their performance across deployment-relevant speech domains.

Read how Natural Language Processing in Action, Second Edition explains word error rate →

Related concepts

These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.