Speech recognition (ASR)
A model that turns recorded or live speech into text, fine-tuned on the customer's own audio, terms, and product names.
Category
How a model is chosen and adapted
Related terms
Speech recognition, or automatic speech recognition (ASR), is a model that turns recorded or live speech into text, so the rest of a system can read a call the way it reads a document.
Why it is fine-tuned
General speech-to-text models struggle with product names, local terms, and the audio quality of a phone line. Fine-tuning on the customer's own audio and vocabulary, inside the customer's environment, is what makes the transcript usable for quality review and compliance checks. As with any small model, the customer's own people grade the test set before training.
What it does not do
It does not identify who is speaking: there is no speaker identification and no voiceprint, because a voice is biometric data. Audio stays in the customer's environment. The contact-center speech use case shows it feeding a small language model that checks every call for quality and compliance; a team lead decides every score that affects a person's review.
Want these terms worked through on one of your own decisions? Start with a Decision Blueprint