Hi,
I am interested in using a chatbot for customer support and am trying to figure out how to evaluate their performance. I am looking for a way to test different chatbots in regards to the ability to understand enquiries and match them with answers.
I know that e.g. Pandorabots provide this as a basic function, but one still has to provide a lot of permutations of each potential enquiry for a specific topic/answer. Other systems like cleverscript seem to have pattern recognitions to reduce the effort, but I wanted to see what the experienced members of this group recommend.
The turing test isn’t that important for what I am looking for as that seems to be focused on the “human likeness”. I am more interested in how to get the question - answer pairs as quickly up an running and with a high accuracy + flexibility ( I know I ask for a lot). I always hoped a great NLP engine would understand “like a human” and make that job easier.
Thanks for the help.
Bora