Predicting the Subjective Responses Emotion in Dialogues with Multi-Task Learning

This repository contains the source code and supplementary materials for the paper:

Predicting the Subjective Responses Emotion in Dialogues with Multi-Task Learning

Anticipating the subjective emotional responses of the user is a critical capability for automatic dialogue systems. In this work, given a piece of dialogue, we address the problem of predicting the subjective emotional response of upcoming utterances—that is, forecasting the emotion that will be expressed by the next speaker.

Repository Contents

• Multi-Task_for_3Emotional_Classes.ipynb
  Notebook for multi-task learning experiments with 3 emotional classes.

• Multi-Task_for_7Emotional_Classes.ipynb
  Notebook for multi-task learning experiments with 7 emotional classes.

• Single-Task_for_3Emotional_Classes.ipynb
  Notebook for single-task learning experiments with 3 emotional classes.

• Single-Task_for_7Emotional_Classes.ipynb
  Notebook for single-task learning experiments with 7 emotional classes.

• Supplementary_Materials.pdf
  Supplementary document providing additional details on the methodology, experimental setup, and results.

Getting Started

Prerequisites

• Python 3.x – Ensure you have Python 3 installed.
• Jupyter Notebook or JupyterLab – Required to run the provided .ipynb files.
• Additional Python libraries as specified within the notebooks.

Installation

1. Clone the Repository:

   git clone <repository-url>
   cd <repository-directory>

2. (Optional) Create and Activate a Virtual Environment:

   python -m venv venv
   source venv/bin/activate  # On Windows, use: venv\Scripts\activate

3. Install Required Packages:

   Install any necessary dependencies using pip. Refer to the notebook cells for specific instructions if needed.

Usage

Open the desired Jupyter Notebook using Jupyter Notebook or JupyterLab. For example, to run the multi-task learning experiment for 3 emotional classes:

jupyter notebook Multi-Task_for_3Emotional_Classes.ipynb

Each notebook is self-contained and includes instructions on how to run the experiments, along with code and comments to guide you through the process.

Citation

If you find this work useful in your research, please consider citing our paper:

@inproceedings{hayat2021predicting,
  title={Predicting the Subjective Responses Emotion in Dialogues with Multi-Task Learning},
  author={Hayat, Hassan and Ventura, Carles and Lapedriza, Agata},
  booktitle={Proceedings of the Conference on Affective Computing and Intelligent Interaction},
  year={2021},
  publisher={Springer}
}

For the complete reference, please refer to the paper available here.

Contact

For any questions or further information regarding this repository, please contact:

Mr. Hassan Hayat
Email: hhassan0@uoc.edu

License

This project is licensed under the MIT License. You are free to use this code for academic and research purposes at your own risk.