What is a speech recognition custom vocabulary?

Definition

A custom vocabulary can include names, acronyms, product terms and preferred spellings that a general speech model may misrecognize. A system may use those entries as model hints, post-transcription replacements or both.

Corrections need context and review because a simple replacement can damage unrelated words that sound similar. Effective dictionaries track real errors, preserve intended capitalization and apply the narrowest reliable rule.

ELI5

A speech recognition custom vocabulary is a list of special words and corrections that helps a dictation system understand terms a general model often gets wrong. It is especially useful for names, acronyms and technical language.

For example, if a model repeatedly writes a product name incorrectly, the app can correct that exact spoken form to the approved spelling. The rule should be tested so it does not replace an ordinary word in the wrong context.

Frequently asked questions

What belongs in a speech recognition custom vocabulary?

Names, acronyms, technical terms, product names and preferred spellings that recur in real dictation can belong in it.

Can custom vocabulary rules create transcription errors?

Yes. Broad or context-free replacements can change unrelated words, so rules should be narrow and tested on representative speech.

Videos explaining speech recognition custom vocabulary

  1. Portrait of Pat Simmons beside the words Build Your Own Dictation App