Out-of-vocabulary (OOV)

The OOV feature improves transcription accuracy for audio that contains uncommon terms, or common terms in uncommon contexts, with custom JSON dictionaries of phrases and their pronunciations

OOV (out-of-vocabulary) is an ASR tuning feature designed to improve transcription accuracy for audio that contains brand- and industry-specific terminology. OOV enhances existing language models with new words and preferential treatment for those words.

OOV's best use cases are

  1. to add phrases that are otherwise unlikely to occur in transcribed audio, and

  2. to add words with no phonetic representation in the model's vocabulary.

Both of these scenarios are prone to transcription errors and can be accounted for with OOV. Common examples include phrases with standard vocabulary words in uncommon contexts; made-up or novel words; and people, company, brand, location, and product names.

OOV may seem similar to Voci's Substitution feature, but it solves transcription errors from a different angle. OOV anticipates errors by using a JSON dictionary supplied with the transcription request to expand the language model's standard vocabulary, so with OOV, the transcription error is never made in the first place.

Substitution corrects errors after they are made, and works best for correcting common transcription errors in particular audio scenarios. OOV works best for detecting nonstandard vocabulary.

Version and language requirements

OOV is a new feature for V-Blaze 7.5 and for V-Cloud 1.9.5.

As of the initial release, only eng-us:callcenter supports OOV, but eventually all Voci language models will support OOV. Any language model pack with version number 7.5 or higher supports OOV.

OOV fully deprecates the Hinting feature, and is supported with all other output tuning features.

OOV supports multilingual transcription.

Workflow

There are two steps to using OOV for transcription:

  1. Develop a dictionary of OOV terms and their pronunciations.

  2. Pass the parameter oov as a stream tag with your transcription request. Set the value of oov to the vocabulary and dictionary JSON.

    For example, this cURL request demonstrates passing an OOV dictionary as a file named example-oov.json:

    curl -F output=text -F oov="</path/to/example-oov.json" -F file=@example.wav https://asr.example.com/transcribe
    Download a copy of the file used in this example: example_oov.json
Learn more: