The idea came from the concept of improving employee quality. The most successful of companies need good support, and sometimes physically managing employees or learning entire customer service booklets can be cumbersome. Creating something to improve upon this aids in better customer assurance and better employee quality of work.
What it does
Borg Support is an automated feedback system for customer service representatives. It reminds the representative of talking points and helps monitor customer satisfaction. Borg Support uses speech recognition to transcribe audio in real-time and monitor customer support calls. We use sentiment analysis to identify potential rough patches in the conversation and alert the customer support representative. A number of key talking points are also displayed to the representative, this may include important points from the script, reminders to upsell, etc.
How we built it
We built our project using Docker to host the backend which has Google's Speech To Text library and python to gather data. On the frontend for accessing the STT data, we used Angular.
Challenges we ran into
Our main challenges were related to communication between the front end and back end. Limitations in the ability to record and send audio from a web-based front end made integration and testing difficult. Our second main issue was in the reliability of speakers diarization. We found that available tools were not able to reliably differentiate between similar-sounding speakers.
Accomplishments that we're proud of
This is one of the first hackathons our group has gone to before (only having gone to 1 hackathon at most if we've ever been to one). Being able to work under pressure like this and having something cohesive enough to submit alone is already an accomplishment for us and something we plan on improving upon and getting more consistent with in the future as we attend more hackathons.
What we learned
We learned how to use Angular in a more in-depth manner and got more familiar with Google API. Speech to text technology was something our team has never worked with, so this was a learning point for all of us.
What's next for Borg Support
We want to add more responsive recommendation functionality, in order to improve the quality of suggestions. We'd like to generate suggested talking points using machine learning based on previous support calls to further improve customer satisfaction. Improved accuracy in real-time transcription would be another goal to aim for. Creating a more appealing, simple-to-manage frontend is also planned.