While caught in the the excitement of coming up with project ideas, we found ourselves forgetting to follow up on action items brought up in the discussion. We felt that it would come in handy to have our own virtual meeting assistant to keep track of our ideas. We moved on to integrate features like automating the process of creating JIRA issues and providing a full transcript for participants to view in retrospect.

What it does

Minutes Made acts as your own personal team assistant during meetings. It takes meeting minutes, creates transcripts, finds key tags and features and automates the process of creating Jira tickets for you.

It works in multiple spoken languages, and uses voice biometrics to identify key speakers.

For security, the data is encrypted locally - and since it is serverless, no sensitive data is exposed.

How we built it

Minutes Made leverages Azure Cognitive Services for to translate between languages, identify speakers from voice patterns, and convert speech to text. It then uses custom natural language processing to parse out key issues. Interactions with slack and Jira are done through STDLIB.

Challenges we ran into

We originally used Python libraries to manually perform the natural language processing, but found they didn't quite meet our demands with accuracy and latency. We found that Azure Cognitive services worked better. However, we did end up developing our own natural language processing algorithms to handle some of the functionality as well (e.g. creating Jira issues) since Azure didn't have everything we wanted.

As the speech conversion is done in real-time, it was necessary for our solution to be extremely performant. We needed an efficient way to store and fetch the chat transcripts. This was a difficult demand to meet, but we managed to rectify our issue with a Redis caching layer to fetch the chat transcripts quickly and persist to disk between sessions.

Accomplishments that we're proud of

This was the first time that we all worked together, and we're glad that we were able to get a solution that actually worked and that we would actually use in real life. We became proficient with technology that we've never seen before and used it to build a nice product and an experience we're all grateful for.

What we learned

This was a great learning experience for understanding cloud biometrics, and speech recognition technologies. We familiarised ourselves with STDLIB, and working with Jira and Slack APIs. Basically, we learned a lot about the technology we used and a lot about each other ❤️!

What's next for Minutes Made

Next we plan to add more integrations to translate more languages and creating Github issues, Salesforce tickets, etc. We could also improve the natural language processing to handle more functions and edge cases. As we're using fairly new tech, there's a lot of room for improvement in the future.

Share this project: