Many of our team members work in startups. Scrum meetings are held very frequently and there are some recurring issues.

  1. There are times where the meetings are not held on time due to forgetfulness
  2. While brainstorming for ideas or solutions, some of our team members' valuable inputs are not recorded down, especially during long meetings.

So we decided to build a voice-enabled meeting assistant.

What it does

A user has to fill up a simple form, providing the phone numbers of her team members and the meeting time. EMMA, or Eureka Moments Meeting Assistant would then call everyone involved in a meeting.

Throughout the meeting, one phone has to remain in the call so that EMMA can track the voices in the meeting. When a person says "Emma", EMMA would be ready to record down the next idea or solution.

At the end of the meeting, an email would be sent to all the team members, containing all the Eureka Moments!

How we built it

We use Nexmo Voice APIs to initiate calls to meeting participants, an Express application hosted on AWS EC2 to serve the meeting creation page and the webhooks for Nexmo. Lastly, we use Google Drive APIs to dynamically create a new Google Document for every bot-created meeting.

Challenges we ran into

Transcribing voice into plain text proved to be a huge challenge in this hackathon. We used various voice transcribers like IBM Watson and Amazon Lex. However, we are unable to achieve 100% accuracy.

Accomplishments that we're proud of

We managed to come with a minimal viable version of EMMA in just 20 hours with the API of a sponsor, Nexmo.

We also explored speech-to-text technologies such as Google Cloud Speech, Amazon Lex chatbot as a Nexmo Websocket call participant (backed by a AWS lambda function) and IBM Watson.

What we learned

Voice transcribing is a hard problem to solve. Although its uses would potentially streamline our workflow immensely, we, as humans living in this world, have not perfected the craft yet.

What's next for EMMA

To bring EMMA to the next level, we would develop our own voice transcription technology to ensure close to 100% accuracy. We would also integrate with various task management tools like Trello, Slack and etc.

We would continue using Nexmo's awesome collaborative and timed phone conversation APIs.

Share this project: