YouTube Comments Road Evaluation

This project analyzes YouTube video comments. The ultimate goal is to help creators make historically accurate, engaging, and audience-aligned videos by creating a data-driven roadmap based on viewer sentiment.

🎯 Project Objective

• Input: A YouTube video link
• Output: A structured roadmap based on clustered negative comments
• Purpose: Detect common complaints from viewers and guide future video production accordingly

🧠 Technologies Used

• YouTube Data API v3: To fetch video comments
• Custom Text Cleaning Functions: For multilingual pre-processing (English + Turkish)
• Sentiment Analysis:
  • Custom-trained English sentiment model
  • Language support for both Turkish and English
• Embedding & Clustering:
  • Sentence embeddings via paraphrase-MiniLM-L6-v2
  • Comment grouping with K-Means Clustering
• Recommendation System:
  • Powered by Gemini (LLM) to produce smart, human-like suggestions based on clustered comment content

🧪 Methodology

1. Comment Extraction

• YouTube comments are retrieved using the YouTube Data API v3.

2. Cleaning & Preprocessing

• All comments are cleaned using language-specific clean_text methods (e.g., punctuation removal, stopwords, etc.).

3. Sentiment Detection

• A custom sentiment model filters out negative comments for further analysis.

4. Vectorization & Clustering

• Remaining negative comments are embedded using SentenceTransformer.
• K-Means clustering groups similar comments to detect common issues.

5. Recommendation Generation

• Gemini LLM analyzes the themes from each cluster and generates improvement suggestions for the content creator.

🔍 Key Use Cases

• Content Feedback Loop: Provide YouTubers with actionable insights from negative feedback.
• Multilingual Support: Process both Turkish and English comments natively.
• Roadmap Creation: Help producers understand what to improve in the next video (e.g., historical accuracy, character development, tone).

📦 Dependencies

To run the project, make sure to install the following Python libraries:

• pip install sentence-transformers
• pip install scikit-learn
• pip install transformers
• pip install langid
• pip install nltk
• pip install stop-words

🚀 Future Work

• Support other languages to broaden reach