Hello everyone, I was wondering what would the best way of building a retrieval based question answering bot? I have a set of 500 questions and answers and I am using it as the knowledge base for the bot for answering service desk questions.
In case of an operational bot (which performs some functions like ordering pizza, providing weather alerts) the intents are kind of limited but when it comes to answering FAQs, each question might have a different intent. Furthermore, we also need to provide at least 5-10 examples of each intent in the data file. This is too much data to prepare for 500 questions. This is what my understanding is please correct me if I am wrong.
It would be great if anybody could tell me if I am going right or if there’s any better way of doing this. (The bot that I am building should be fully on premise)