When trying to use these utterances the bot takes the confirm status as none instead of confirmed.
Example:
"confirmationStatus": "Confirmed", "sourceLexNLUIntentInterpretation": null }, "inputTranscript": "yes" }
"confirmationStatus": "None", "sourceLexNLUIntentInterpretation": null }, "inputTranscript": "yes it is" }
Has anyone ever faced this with AWS Lex?
Tried to clean up any slots that would expect "yes it is" or similar utterances.
Expected:
"confirmationStatus": "Confirmed", "sourceLexNLUIntentInterpretation": null }, "inputTranscript": "yes" }
Actual:
"confirmationStatus": "None", "sourceLexNLUIntentInterpretation": null }, "inputTranscript": "yes it is" }
Amazon suggests using confirmIntent or confirmSlot for yes/no questions. But because of the limitations of the input recognition, I chose to use a slot instead.
I created a custom slotType that resolves values to either "Yes" or "No" and uses a long list of synonyms for each, to capture the many ways to answer. It allowed me to also catch thumbs up/down emojis, as well as just "y" or "n" for lazy users.
Users do tend to throw more colloquial phrases at bots in an attempt to trip them up, so I also added phrases like, "you bet", "nailed it", "not a chance in hell", etc.