We were brainstorming and had a whole bunch of ideas coming up, but none of us kept notes of the ideas. Due to this, we built this easy-to-use app that helps people keep meeting minutes and compile summaries with it.
What it does
It records an audio file of a meeting/discussion, and then uploads the file to our backend service, which utilizes Watson and also an homebrewed NLP clustering method that gives us the text and summaries of the meeting. On the app, users can use drag and drop different sentences to organize there meeting minutes.
How we built it
We used expo as our mobile front-end, Flask as the back-end server hosted on GCP, and used IBM Watson to convert speech to text. We also developed an NLP clustering algorithm with Scikit-Learn and Gensim, giving rough topic grouping of the context.
Challenges we ran into
It was hard to find the best speech-to-text API, we chose Watson over others since it had the best sentence splitting performance. We also had a hard time building a good clustering method since we had little datasets to train our model.
Accomplishments that we're proud of
We utilized machine learning methods to solve a challeng.
What's next for Minute Maid
We hope that Minute Maid can become a meeting agent that not only takes notes but also helps people to conduct meetings easier, such as keeping records of regular meetings, meeting attendee tracking, minute sharing etc. Few of the minute keepers use AI approaches to solve problems, and we hope that Minute Maid can be exploit ML to tackle these challenges.