README.md

metadata

license: apache-2.0
task_categories:
  - summarization
language:
  - en

📚 Clip-CC Dataset

Clip-CC is a high-quality dataset of 200 short educational videos (1 minute 30 seconds each) paired with human-written summaries. It’s designed for tasks like video summarization, video-language alignment, and multimodal learning.

---

🧾 Dataset Summary

• 🎞️ 200 videos: Named 1.mp4 to 200.mp4
• 📝 Each video has a concise human-written summary
• 📂 File references stored in a JSONL file (metadata.jsonl) with paths to each video

---

💡 Use Cases

• Video summarization and caption generation
• Vision-Language alignment
• Video QA and downstream educational AI tasks
• Fine-tuning multimodal models (e.g., Flamingo, Video-BERT, LLaVA)

---

📦 Dataset Structure

Each entry has:

Field	Description
id	Unique ID, e.g., 001
file_name	Relative path to video file
summary	Human-written summary of video content

🔍 Example

{
  "id": "001",
  "file_name": "clips/1.mp4",
  "summary": "An introduction to gravitational force using animations."
}