Created on 2nd March 2024
•
Streamlined Post-Meeting Review: The inability to identify and summarize the topics discussed during a meeting review process. With this product users can quickly refer back to specific segments of the conversation without having to sift through the entire meeting transcript, and analyse sentiments, and other metrics, saving time and enhancing productivity.
Licensing educational videos/podcasts: In todays world creators don't have the ability to monetize on reshare or reclip of videos and Sentient Scribe AI solves this by using Story Protocol to create an NFT of the video metadata to license and monetize it appropriately.
Content Assimilation: Content creators and viewers don't have a method to assimilate and organize their content that they have posted throughout their creator phase, this product allows them to analyse their content by storing them in Ceramic as a graph.
Enhanced Meeting Accessibility: By accurately mapping transcriptions to the correct speaker, the assistant ensures that meeting minutes are not only precise but also accessible.
Emotional Intelligence Insights: Tracking the sentiments of different sentences provides a layer of emotional intelligence to the meeting analysis. This can help managers and team members understand the tone and mood of the discussion, identify points of contention, and recognize moments of agreement or enthusiasm, fostering a more empathetic workplace.
Developing Sentient Scribe AI was an ambitious project that presented several unique challenges. Below are the key hurdles we encountered, along with the strategies we employed to overcome them:
Integration with Livepeer SDK
Challenge: We faced difficulties in using the livepeer-sdk based on their quick onboarding process mentioned in the documentation, the functions that were mentioned in the SDK for upload were not available in the SDK as shown in the documenation
Solution: We had to find different workarounds to call the API and work for the upload functions and create a pooling system once uploaded.
Challenge: Streaming hls format video in javascript did not work out of the box, based on the documentation.
Solution: We had to write our custom video player to stream the hls format video provided by Livepeer.
Challenge: Creating sentiment analysis tags for the sentences proved difficult by using out of the box model weights.
Solution: We had to finetune the model in order to get the right sentiments for each sentences.
Challenge: Fluence sdk built scripts kept crashing out of the box and had to be manually edited to make it work on a MacOS system
Solution: We had to run a local fluence node instead of a DAR node to create our aqua scripts and after doing that we were able to test the ceramic db functions
Tracks Applied (3)
Fluence
Livepeer
Story Protocol
Cheering for a project means supporting a project you like with as little as 0.0025 ETH. Right now, you can Cheer using ETH on Arbitrum, Optimism and Base.