We are group 31 and we were very interested in the challenge TheHutGroup proposed in their brief presentation, so we decided to apply for that and for Bloomberg's challenge for best tech. The objective was to provide a way to sort emails according to customers' queries to their 'Customer Support' department using Natural Language Processing techniques. We immediately dived into the challenge.

For this purpose we have created a webpage that displays e-mails and allows filtering based on tags such as "Delivery", "Returns" and "Feedback". In order to get a such tags we scanned the e-mails using Dialogflow, a tool that lets you train your own bots for intent recognition.

The on-line platform consists of a webpage with the sorting filters on the left with access to a form to submit an e-mail to the system for testing and demonstration purposes. It also allows to mark inquiries as "Resolved" so they are in a separate area without disturbing.

It should be noted that we have been working on a Machine Learning model to filter SPAM before passing the e-mails onto the node.js server. Albeit, we have not been able to finish it due to time constraints, even though we plan to complete it in the near future.

Share this project: