为了找出我正在用句子写的情绪，我该如何使我的聊天机器人成为现实？

Question

I'm trying to make a chatbot to ask about user's feeling on exhibition.

(with IBM watson-conversation workspace on web.)

First, question is

It has 7 steps asking how user's feel on exhibition.

I can make this flow with direct answer for the question.

However It doesn't work if user's answer was not what I expect.

It just goes back to the beginning.

How can I make a dialog flow to not going back steps?

Secondly

In these steps, I want to know how user feels at the sentence.

If I want to know that feelings at the same time, how do I make those?

I declare those word into entities, I can see them on the testing chatbot section.

Thank you.

Answer 1

For Question 1:

Conversation service works like that. If it doesn't find the matching intent in the continuing flow, then it looks for that intent in the root node. After that also if it doesn't find the intent, then it goes to the anything_else intent.

One simple solution to your query is to give your users some options to choose from if it's possible. That way you'll have more control over what user can enter and simultaneously you can maintain your flow.

And if that's not possible, then try to include all possible intents in the flow, where all sorts of variation from user text can trigger the intent and the flow is maintained.

For Question 2:

I believe you are asking about using Tone Analyzer service to identify user emotions for the texts that the user entered and you probably want to direct the user to a specific flow in Conversation depending upon what emotion was detected.

Keep one thing in mind that all Watson services are independent of each other, that means one service can't identify any other services that's connected to the same application.

So when user enter the text, first give that text to Tone Analyzer from the application and once you get the response, parse the emotions and give it accordingly to Conversation service, so that proper dialog can be detected.