I am using amazon lex in my connect flow. It takes an audio input and detects the intent accordingly. I want to do some form of preprocessing of the transcripted text before detecting the intent for it (because the preprocessing might change some words in the original text, and hence, change the resulted intent). So, ideally I would like to be able to capture the text before the intent gets detected. This means that using lambdas within an intent (codehook or fulfillment) is not exactly what I need to do. I need to (1) capture the transcripted text (2) perform preprocessing on it (3) send the preprocessed text to lex and let it do its magic.
I can see a similar preprocessing layer explained in this link but it assumes that the text is not audio based (the link is a tutorial for integration with twilio). So, what is the best way to do that?
