Sestek SR translates everything a speaker says into text. Denormalization is not applied to the output of speech recognition. The reason for that is the fact that the context of speech inside the input is unknown. Applying denormalization to speech recognition in a dialogue of which the context is unknown would corrupt the speech recognition.

 For example, let’s assume the question you ask the user on your application is as follows; 

‘Kaç Euro transfer etmek istiyorsunuz?’’

The user responds:


Because the Sestek SR would not be able to distinguish whether the question being asked to the user is a financial transaction or not, the final output text would be as below;

‘’ yüz’’ 

Related Products: Conversational IVR 1 (NLP+SR), Conversational IVR 2 (NLP+SR+Orhestrator)