We go through a lot of meetings, at school or in the workplace; in fact, more than 25 million meetings occur every day in the US, but most of these are not very productive. People would be too focused on taking notes to remember what matters, such as understanding what other team members are saying or actively contribute to the conversation. Also, it is likely that valuable ideas are lost because no one took adequately detailed notes. Meetings could be much more productive if there was a web app to automatically take notes, and it reduces the need for repeat or follow-up meetings as well. In the US, more than $37 million is spent on unproductive meetings per year, and we made it our mission to change that.
What it does
Autonote is able to document everything that is said in a meeting and summarize the main ideas. Hence, allowing humans to focus on creating ideas rather than making sure everything is recorded.
How we built it
The meeting transcript was recorded with a unique implementation of the IBM speech-to-text API. We used react to build a beautiful frontend and nodejs to run the local server. We formulated a summary algorithm that can pick out the key ideas from a meeting.
Challenges we ran into
It was hard to integrate the IBM speech-to-text API and find an API or algorithm that would summarize the text. Also, we did not get a lot of sleep (and spent too much time on Hackenger Hunt…)
Accomplishments that we're proud of
We were able to implement our own text summarization algorithm which is a step towards making a fully functional natural language processing system.
What we learned
What's next for Autonote
- allow users to change the names of “Speaker 1” or “Speaker 2”
- Add an option for how “condensed” they want the text to be (½ of the original length?)
- Allow users to make an account and save all their past meeting info on it