Diagnostics

The diagnostics center is useful for identifying recognition issues that may occur in your project. To use the diagnostics center, simply ask a question and the diagnostics will show when you have opened the Diagnostics view. You can find this in the upper right corner. When clicked, the Diagnostics screen will be displayed and remain visible throughout the entire testing process.

'Knowledge', 'Intents', and 'Entities' are integrated into the NLU-flow overview in the upper-left corner. NLU stands for 'Natural Language Understanding'. The NLU flow is the process in which the customer's query is analyzed, assessed, and answered.

Intent Recognition

This shows which intent was matched to the question along with its threshold and score. If you're using intents in your project, and you do not see a matched intent, then you might consider creating one to match with the question that was asked - or add the question to an existing intent and retraining your intent model.

Rule-based Recognition

The detected entities and keywords and which entities matched with the question. If the wrong entities are matching with the question, check that your entities are not overlapping. If no entities are matching, you might consider adding one of the keywords to an existing entity, or creating a new entity so that there is a match.
Select the 'view' icon to view the words in the matched entity.

Generative AI

This shows the GenAI Response and which resources were used to provide the output.

<>API Response

This shows which Q&A matches the question received, language detection and translation (if enabled for your project), and context and session value results. You can go directly to the Q&A by selecting the 'Edit Q&A' button on the matched Q&A.

More help with troubleshooting issues:

If you're struggling with resolving recognition issues, you can always ask Support for help. Please send us the following information in your ticket:

• The url of your project
• The question asked
• The answer received (with the Q&A article ID)
• The answer you would expect it to receive (with the Q&A article ID)