We created a natural language processing model to transcript audio instruction into text.

Broadly trained models can be used, with a degree of transfer learning, but a custom approach specific to the client lexicon was required for the system to understand the nuances of speech.