Domain specific Chatbot with limited Data

 

In recent years, chatbot and its underlying technology has become popular among tech and business community. Chatbot has evolved from answering redundant questions to answering domain specific inquiry with context awareness. Future ask from chatbot would be: seamlessly digest information and provide an answer in actionable format. Above process gets more complicated when we use it for a new domain where terminology doesn't always follow a natural language framework and the corpus is full of abbreviations or lack of data abundance or users have very limited understanding of the domain, In these scenarios, training an AI bot becomes more complex at multiple levels. In a real-world scenario there are only few SME- Subject Matter Experts who can act as reference points, areas such as customer solutions, maintenance, legal and medical are continuously growing and new topics emerge almost every year. Solving such a moving target using AI would be challenging. This talk I would like to present my learnings on building domain specific chatbot, which includes, how to start with no data, develop ontology, work with limited Q&A samples, use ML tools-incorporate bot framework, deploy& gather feedback and retrain.

 

Who is this presentation for?

Data scientists & Product managers

 

Prerequisite knowledge:

basic data science, product managers

 

What you'll learn?

 What you need to know before one builds a chat bot, business& technology