Intent Identification Issue in Kore 10.7 Dialog Task

AI for Service
1

Reaching for Kore 10.7 help from Kore.ai Engineering Team.

Context & Issue: Building intent identification leveraging NLU of 10.7 with manual few-shot training of utterances. Disambiguation and fallback flow trigger when user utterances are ambiguous to get the intent right. It is not built to server the intent.

Question: To maintain operational reliability, we designed one dialog task for all intent identification instead of individual dialog tasks. This causes an issue where the identified intent name from NLU cannot be passed to downstream Kore 10.7 Bot and/or systems with the “intent name” from NLU; only the Dialog Task name is passed.

Example: Need to pass the intent identified name “intent_1” from NLU engine of Kore 10.7 in single dialog tasks “dt_dialog_task” (where all intent identification happens in the same dialog task) to the Genesys system as “intent_1” not the “dt_dialog_task”.

Platform: Kore 10.7 Bot

Bot Type: Standard

2

hi @kumar_chinnakali `

Recommended Practice in Kore.ai 10.7

The standard and recommended approach is to train every Dialog Task with its own set of utterances. This allows the NLU to directly identify and trigger the correct dialog without additional overhead.

To ensure operational reliability and simplify maintenance, please follow the standard process of training each Dialog Task with utterances rather than centralizing all intent detection in one dialog.