Hi All
Not getting proper response from bot
User: What is your name?
Bot: Your name is Xyz
user: please confirm your email Id?
Bot: Email id is xyz@gmail.com
when I’m asking What is your name? . Bot is answering Email Id .instead of name
please guide me on this
I would recommend you remove all but the first Intent node. The Intent node guides you to the desired dialog flow so your design is not configured correctly. Remove all but “Test Inbounds” and I would expect it to work correctly.
@john.nicholson I have removed all intent and added all Utterances in test inbound but still its replying with Email ID and Not Name
If your bot is still not preforming as expected, I suggest you use the debug tools to better understand what the bot is doing during execution. When you start talking to the bot, be sure to click on the ‘bug’ icon at the top of the chat window.
@sachin I am confused about what you are trying to achieve. Your initial comment has the user interrogating the bot, rather than the bot asking questions and parsing the responses.
If that is what you want - the user asking the bot questions, then a dialog like this is not the way to approach it. Instead that would be a situation more akin to FAQs, which on the Kore Platform are handled through the Knowledge Graph.
A dialog is more suited to the bot asking the user a sequence of questions to gather supplemental data about the intent.
Perhaps you can clarify your use case?
The question I always ask people in situations like this is what, as a human, is it about that utterance that makes you believe that the correct understanding is all types of calls?
NL training a bot is a reactive process, the user is in control and can say anything. It is your job as a bot developer to predict what they might say and add the appropriate training.
Now a normal List of Items entity is about single classification, it’s job is to pick the best of the (in this case) 4 choices. Those 4 are what is expected, so one of those 4 it must be. In this situation the training for “I call” matches “… my calls”, which is what I would expect.
So if you want a different answer then you need to provide additional choices to the entity. In all kinds of NLP situations it is very difficult for a system to say “none of the above” unless there is an explicit training pot providing a viable alternative.
