Hi, i have couple of questions which i hope will give bit more clarity on dialog engine bot flows:
1. How is context mantained in the bot flow?
2. One use case iam looking at is if there is no intent match, then capture the request the customer made (for simplicity sake, assume this is a chat) so that it can be presented to the agent. Is that possible?
3. A situation i came across was, if i type 'i want to fly an airplane' , i would have expected the match to be zero and a no match being triggered, however, it matched an intent i had configured (with a low confidence e.g 30) with the bot asking me to verify if 'i meant xxx'. I would have though that common words such as 'i want to' would be a stop word and would be ignored and the match would be zero since i dont have any other intents with the utterance containing fly or airplane. What can be done in this situation? Is there a way to direct the NLU to have a lower confidence?
4. Is there a way to overide the 'no match' defaults so that the logic that resides in the 'no intent' path for 'ask for intent' node takes effect. A use case for this is to send the customer to an agent instead of asking them to re-enter their request.
5. What is the best way to handle end of conversations. Example use case is, if an intent is detected and a response is provided, we ask the customer if there is anything else , and they respond with something like 'nah, iam good', if there are spelling mistakes, what happens is that it tries to confirm if the intent set for handling end of conversations is a match which is not ideal experience. Is it better to handle this with a 'no intent' match however, i can hit the issue mentioned in 3 above?
#ArchitectureandDesign