Vibescope

Inspiration

The inspiration behind Vibescope came from the recent removal of the ability to view dislikes on YouTube videos. This change made it challenging for users to identify low-quality videos based on community feedback. To address this issue, we decided to leverage sentiment analysis on the comments of YouTube videos to detect how commenters reacted to the content and generate an approval ratio for the video.

What it does

Vibescope is a website that utilizes Hume AI to perform sentiment analysis on YouTube comments. By analyzing the sentiments expressed in the comments, Vibescope provides an overall understanding of how viewers are reacting to a particular video. The system classifies commenters' reactions as positive or negative, allowing users to get a sense of the video's quality based on community feedback.

How we built it

We employed a variety of technologies to build Vibescope:

1. Hume AI is a powerful tool for working with human emotion with technology. Hume AI provided the necessary algorithms and capabilities to analyze the sentiments expressed in the comments accurately. We built a server using Python and Flask to access Hume.
2. We used Google's YouTube API to access comment data for specific videos. We built a server in Node.js hosted on AWS Lambda to access the YouTube API.
3. We developed a user-friendly website interface in Svelte that allows users to enter the URL of a YouTube video and obtain sentiment analysis results for the associated comments.

Challenges we ran into

Performing sentiment analysis on each comment created too much latency and didn't scale well. So, we combined all the comments into a "conversation" and had Hume analyze all of it at once. Then, we used the weights for each emotion to create a breakdown of approval and disapproval.

Accomplishments that we're proud of

We were originally using the Node.js serverless application to access both the YouTube and Hume APIs. We struggled to use Hume in Node.js and we found that they have a simpler integration in Python, so we created a separate server using Python and Flask. The website accesses the Node.js server which accesses the Flask endpoint, then all the data is sent back to the frontend to show the user. We are proud of working past this problem and creating a working solution.

What we learned

Throughout the development process, we learned the importance of community feedback, the power of AI for sentiment analysis, and collaboration skills. YouTube's removal of the dislike feature eliminated a useful metric for users to evaluate videos before watching them. Hume AI showed us that technology does not have to be emotionless. Working as a team allowed us to develop our collaboration and problem-solving skills. We learned to communicate effectively, divide tasks efficiently, and overcome challenges together.

What's next for Vibescope

We are planning to create a Chrome extension for Vibescope to embed approval ratings directly into the YouTube website for each recommended video on the user's page. This way, users don't have to navigate to our separate website to check videos. Instead, they see it within YouTube itself.

Built With

• amazon-web-services
• flask
• humeai
• node.js
• svelte