Watson Conversation Doesn't Recognize New Entities

0
votes

When adding training examples to an intent, it seems the intents don't have the capability to match against untrained entities.

Given the following intent training:

enter image description here

If I test out my intent, I match the intent, but I want to detect the entity even if it's not trained in the intent training set. Here are my results for an untrained entity.

enter image description here

Notice the example test case is identical to the first training user example. The only difference is 'grapes'. Can Watson conversation still provide the unmatched entity 'grapes' so my dialog can continue without re-prompting?

Is there a way to provide a variable in the user example training to match strings?

Why am I asking for this? The user can virtually ask for anything, so there isn't a finite set of entities.

1
watson-conversation

1 Answers

0
votes

If you wanna use just Conversation service, you need to train the Intent #WhatPrices using examples for asking prices and create entities with the correspondent values for Watson Conversation recognize inside the Conversation.

But, if you wanna to detect this words even if it's not trained in the intent training set, you can use AlchemyLanguage API's to detect the keyword in each phrase typed or use NLU to get the keywords, see the Demo example.

For example:

When I typed on NLU demo:

  • "What is the price for two banana?"

They recognize the keyword: bread and price. And you can use the output (JSON) and do something in your application, all depends on your business role.

Obs.: Maybe this information can help you. Probably have others forms but one IBM professional will answer if have.