Working on a proof of concept to extract email id from speech utterances. The expectation is that the utterance will be answer to a question like “What’s your email id?”, hence the response will most likely be short.
I would also already have the email ids (from a database). So I am would be looking for known email addresses. This could be a considered a classification case (?)
Tried doing this using google speech to text APIs but they are very poor at this, thinking if training something specifically would help or not.
Should I go ahead with deepsearch would I need to create my own training data for extracting emails?