Emotion-Based Multimodal Analysis of Portuguese Political Debates - LBMP 2024/2025

This repository contains the code to process audio files, images and text regarding 2024 Portuguese Political Debates.

---

Repository Structure

├── audio_files/
│     ├── ad-be.wav        # Important !!!
│     ├── ...
│      
├── emonet/
│     ├── ...
│
├── Video_Frames/
│     ├── ...
│
├── audio_analysis.ipynb
│
├── audio_processer_2.py
│
├── clustering.ipynb
│
├── frames.ipynb
│
├── README.md              # Important !!!
├── text.ipynb
│
├── total_and_analysis.ipynb
│
└── requirements.txt       # Important !!!

---

Regarding Audio

• To run the audio model, the audio_files folder should have all debates audio files.

• audio_processer_2.py: File with the pre-trained model. Analyses all audio files from audio_files folder and creates 1 csv per debate in the SER-Odyssey/emotion_csvs/30secs/ folder.

• audio_analysis: Notebook that creates plots to analyse debated and political parties based exclusively on the audio generated csv's.

---

Regarding Image

• frames.ipynb: This notebook runs EmoNet and analyzes the files in Video_Frames. Users should input the debate they wish to analyze in the first cell.

---

Regarding Text

• text.ipynb: This file processes the text data associated with the debates.

---

Regarding Analysis

• total_and_analysis.ipynb: This notebook combines the audio and image data, performing analysis and clustering.

• clustering.ipynb: This notebook performs clustering based on the extracted features.

---

Additional Information

• requirements.txt: Lists Python dependencies required to run the project.