YoutubeToxicMeter

• 

  YoutubeToxicMeter Homepage

Inspiration

As viewers, it can be challenging to grasp the overall sentiment of a video without scrolling through hundreds of comments. And for content creators, sifting through an overwhelming number of comments to extract meaningful feedback is nearly impossible. YoutubeToxicMeter solves both of these problems by leveraging AI to analyze and summarize audience reactions efficiently.

What it does

YoutubeToxicMeter allows users to paste a YouTube video link and instantly analyze its comment section. The AI-powered tool determines the distribution of positive and negative comments, providing a visual sentiment breakdown.

How we built it

Our team developed YoutubeToxicMeter using React for the frontend and Python for the backend. We integrated the YouTube Data API to fetch comments and video details, while the OpenAI API was used for sentiment analysis and comment ranking. The model was fine-tuned with a large dataset to enhance accuracy and reliability.

Challenges we ran into

One major challenge was handling the high variability in comment language—ranging from sarcasm to slang—which made sentiment classification difficult. We also had to optimize API usage to process large volumes of comments efficiently without exceeding rate limits.

Accomplishments that we're proud of

Achieving nearly 80% accuracy in sentiment classification through rigorous fine-tuning of hyperparameters. Successfully integrating multiple APIs to create a seamless user experience. Developing an intuitive interface that makes AI-powered sentiment analysis accessible to all users.

What we learned

We gained hands-on experience in natural language processing (NLP), particularly in fine-tuning models for real-world text analysis. We also deepened our understanding of API rate limits, optimization strategies, and the importance of designing a user-friendly interface.

What's next for YoutubeToxicMeter

Real-time analysis: Enhancing speed to provide instant sentiment breakdowns. More nuanced sentiment categories: Beyond positive/negative, incorporating neutral, sarcastic, or spam detection. Customization for creators: Allowing YouTubers to filter comments based on themes (e.g., constructive criticism, praise, questions). Multilingual support: Expanding sentiment analysis to include comments in different languages.

Built With

• openai
• python
• react
• toxiccommentsapi
• youtubeapi