Hi, we are using Amazon Lex with PureCloud in an inbound call flow to recognize/capture the last name pronounced by the caller.
We are using in Lex the slot type AMAZON.US_LAST_NAME for this purpose. Of course, Lex fails detecting some last names but this is an issue the we need to deal with Amazon, however I have noticed that with several common last names when testing them directly in the Amazon Lex Console I have no problem however when I test them using a call flow and the Lex integration I get complete different results.
For example, the following last names (capital letters) works fine (are detected with no problem) testing them directly in the Lex Console, however when using the PureCloud integration I get nonsense:
PATTERSON => rutter, potter, watters
DRAKE => bates, ralph, bass
ANDERSON => underhill, heflin
RAMIREZ => mireles, brackett
VENTURA => spain, tuttle
OROZCO => matos, briscoe, otto
RIVERA => huerta
FULLER => houle
BROWN => towne,
KEYS => heath
Eventually, when trying many times in PureCloud I get the correct result.
I wonder what can be causing this huge difference in results when testing directly with Amazon Lex and when testing in PureCloud.
Could be a telephony issue?
Would it be better to try with Google Dialogflow for this kind of voice bots.
Maybe someone else has had this kind of experience with Lex.
Thank you.