Watson Conversation as an alternative to Watson NLC?

Question

Can Conversation be used as an alternative to NLC? We tried using the same corpus and class names as the intent. We compared the results from NLC and Conversation. The Conversation gave 25% error for a document that was rightly classified by NLC Has anyone tried this approach earlier? Any help is welcome

Answer 1

Neither Conversation or NLC are document classifiers. They are designed to classify questions to a particular intent.

If you are using as it is intended then you can use NLC or Conversation.

They have different learning models. Conversation allows you to customise your model with entities and complex dialog trees.

---

To add to this, Conversation and NLC use two very different scoring methods for confidence. NLC is Relative scoring, while Conversation is absolute scoring. More details here.

https://www.ibm.com/watson/developercloud/doc/conversation/irrelevant_utterance.html