Emotion Detection in Conversational Speech

Project Overview

In this project, we introduce an innovative approach to emotion detection in conversational speech. Utilizing a multi-modal framework that integrates both audio and textual data, our system is capable of discerning emotions within multi-speaker dialogues. This is achieved through the extraction of audio features using pre-trained models and processing of text embeddings, coupled with an attentive bi-directional GRU network. This network dynamically captures the context and the inter-speaker emotional influences.

Key Features

• Multi-modal emotion recognition leveraging audio and text data.
• Utilization of pre-trained models for robust audio feature extraction.
• Implementation of an attentive bi-directional GRU for contextual understanding.
• Evaluation on the MELD dataset, demonstrating effective emotion detection.

Getting Started

These instructions will get you a copy of the project up and running on your local machine for development and testing purposes.

Prerequisites

• Python 3.x
• Pip (Python Package Installer)

Installation

1. Clone the repository git clone https://github.com/siddhantpathakk/emotion-rnn.git
2. Navigate to the project directory: cd emotion-rnn
3. Install the required packages (preferably using Conda/Miniconda): conda env myenv --file environment.yml

Usage

To see the training notebook, please refer to ./audio/train.ipynb

Dataset

This project uses the MELD conversational dataset for training and testing. Ensure that you have the dataset downloaded and placed in the audio directory. We used preprocessed features available here.

Structure

• audio/: Contains audio data and extracted features.
• src/: Source code for the project including model definitions and training scripts.
  • inference.py: Script for performing inference.
  • model.py: Defines the bi-directional GRU model.
  • dataloader.py: Code for loading and preprocessing data.
  • attention.py: Implementation of the attention mechanisms.
• train.ipynb: Jupyter notebook for training the model.
• infer.ipynb: Jupyter notebook for using the model for
• exploratory_data_analysis.ipynb: Jupyter notebook for performing standard EDA about the MELD dataset.
• environment.yml: Conda environment file for setting up the Python environment.
• LICENSE: The license under which this project is distributed.
• README.md: This file, describing the project and how to use it.

Contributing

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

Authors

• Siddhant Pathak - Initial work - @siddhantpathakk

License

This project is licensed under the MIT License - see the LICENSE file for details.

Conclusion

Our methodology opens new avenues in speech emotion recognition by focusing on the nuances of conversational context and speaker interactions. By implementing dual attention mechanisms and a bi-directional GRU, our system adeptly identifies emotional cues