Inspiration
We've read about the Georgia Tech Teacher who created a chat bot to assist him answering all those repetitive questions from his students. In a single semester, his bot answered more than 10 000 questions.
Since we deal with Slack everyday, we know that a whole lot of identical questions are asked everyday throughout the company, taking a lot of precious human time to answer them.
What it does
Our bot can answer basic questions asked anywhere on slack, allowing people to focus on what's important instead of answering those questions.
We automated the learning process and made it so easy for you to train him, that you will never answer another boring question ever again.
How we built it
Using Azure Search, a search service built on Elastic Search, our questions mapping is precise yet flexible. We implemented our bot itself using an Azure Webjob.
Challenges we ran into
Balancing our search result between flexibility and precision. We had to make it flexible enough so it can understand related questions, but also precise enough so he doesn't give you a wrong answer.
Keeping a state in a chat between a user and our system was quite a challenge by itself.
Accomplishments that we're proud of
Question detection quality, our unique learning cycle.
What we learned
Search is hard. Natural language processing is even harder.
What's next for Slack-Moi-Patience
Intercept the question before it is even posted on a channel, create links between questions so we can suggest answer and more !
Log in or sign up for Devpost to join the conversation.