Inspiration

It's easy to zone off in online meetings/lectures, and it's difficult to rewind without losing focus at the moment. It could also be disrespectful to others if you expose the fact that you weren't paying attention. Wouldn't it be nice if we can just quickly skim through a list of keywords to immediately see what happened?

What it does

Rewind is an intelligent, collaborative and interactive web canvas with built in voice chat that maintains a list of live-updated keywords that summarize the voice chat history. You can see timestamps of the keywords and click on them to reveal the actual transcribed text.

How we built it

Communications: WebRTC, WebSockets, HTTPS

We used WebRTC, a peer-to-peer protocol to connect the users though a voice channel, and we used websockets to update the web pages dynamically, so the users would get instant feedback for others actions. Additionally, a web server is used to maintain stateful information.

For summarization and live transcript generation, we used Google Cloud APIs, including natural language processing as well as voice recognition.

Audio transcription and summary: Google Cloud Speech (live transcription) and natural language APIs (for summarization)

Challenges we ran into

There are many challenges that we ran into when we tried to bring this project to reality. For the backend development, one of the most challenging problems was getting WebRTC to work on both the backend and the frontend. We spent more than 18 hours on it to come to a working prototype. In addition, the frontend development was also full of challenges. The design and implementation of the canvas involved many trial and errors and the history rewinding page was also time-consuming. Overall, most components of the project took the combined effort of everyone on the team and we have learned a lot from this experience.

Accomplishments that we're proud of

Despite all the challenges we ran into, we were able to have a working product with many different features. Although the final product is by no means perfect, we had fun working on it utilizing every bit of intelligence we had. We were proud to have learned many new tools and get through all the bugs!

What we learned

For the backend, the main thing we learned was how to use WebRTC, which includes client negotiations and management. We also learned how to use Google Cloud Platform in a Python backend and integrate it with the websockets. As for the frontend, we learned to use various javascript elements to help develop interactive client webapp. We also learned event delegation in javascript to help with an essential component of the history page of the frontend.

What's next for Rewind

We imagined a mini dashboard that also shows other live-updated information, such as the sentiment, summary of the entire meeting, as well as the ability to examine information on a particular user.

Share this project:

Updates