MECP: Early-stage Controversy Detection and Popularity forecasting

📌 Overview

MECP is an early-stage multimodal framework that jointly models controversy detection and popularity prediction via a video-guided viewpoint alignment module and a task-aware Mixture-of-Experts component.

🧷 Dataset

We construct the first English-language short-video dataset TMCD, with 3,061 videos covering 244 real-world topics, annotated with both controversy and popularity scores.

We evaluate our method for event detection on the TMCD and MMCD datasets.

🚀 Quick Start

Step1: Install the dependencies using pip

Our dependencies:

Python: 3.10.6
torch==2.1.0+cu121

Environment requirement:

pip install -r requirements.txt

Step2: Start training and testing

python main.py

📃 Citation

Please cite our repository if you use MECP in your work.