Inspiration
I used to work at a call center. The biggest problem we have is call volume. Having a chatbot would greatly reduce a call center operator's workload and stress. We decided to develop a chatbot to be the first contact for vacation inquiries.
What it does
Handles common vacation inquiries and assess if escalation to a human operator is required. It uses various machine learning models in the backend to generate replies and assess if escalation is required.
How we built it
Front End
- React (cra)
- css, Bulma, animate.css
Back End
- Flask
- seq2seq
Challenges we ran into
Overfitting
Our model overfit the data and produced predictable results. We tried additional models to augment our chatbot and provide a better response.
Deployment
App worked locally but failed on Azure. After consulting documentation and asking around, managed to locate logging information on Azure portal and realized a dependency was missing for the flask application.Ultimately we did not manage to deploy our model on azure.
What I learned
- flexbox, css animation, responsive design
Log in or sign up for Devpost to join the conversation.