This repository contains the code used in creating the Dialogs Re-enacted Across Languages (DRAL) corpus, computing prosodic features of utterances, and modeling the mappings of English and Spanish prosody.

Download the DRAL corpus releases from the releases page.

Note: Releases are incremental. To get the latest release, download each release in sequence, merging the files of each release into the previous.

The DRAL protocol is described in our technical report. For post-processing, see the DRAL-corpus subdirectory.

The code is separated into the following subdirectories. Each subdirectory contains a README.

• DRAL-corpus - Code for DRAL post-processing and preparing the data for feature computation. See the included workflow diagram.
• feature-computation - Code for computing DRAL conversation fragments prosodic features and prosody representations, and for plotting and validating features.
• modeling - Code for DRAL English and Spanish prosody mapping models, baselines, metrics, and feature correlations.

1. Install MATLAB R2023a.
2. Install the MATLAB add-ons Signal Processing Toolbox and Statistics and Machine Learning Toolbox. To install add-ons from MATLAB, select Home > Add-Ons > Get Add-Ons.
3. Clone the Mid-level Prosodic Features Toolkit (February 9, 2023) repository.
4. Before running any of the project MATLAB functions, add the Mid-level Prosodic Features Toolkit to MATLAB's search path. To add a directory from MATLAB, use the built-in addpath function. Alternatively, right-click the directory in the Current Folder view, then select Add to Path > Selected Folders and Subfolders.

1. Create a Python 3.10.4 virtual environment, e.g., with conda.
2. Activate the virtual environment, navigate to the project root directory, then install the required Python modules with pip: pip install -r requirements.txt.
3. Clone the REAPER repository and compile it following its instructions, then add it to your PATH environment variable.
4. Install SoX. Depending on your operating system and install method, you may need to add it your PATH environment variable.

Cite as: Avila, J.E., Ward, N.G. (2023) Towards Cross-Language Prosody Transfer for Dialog. Proc. INTERSPEECH 2023, 2143-2147, doi: 10.21437/Interspeech.2023-1152

@inproceedings{avila23_interspeech,
  author={Jonathan E. Avila and Nigel G. Ward},
  title={{Towards Cross-Language Prosody Transfer for Dialog}},
  year=2023,
  booktitle={Proc. INTERSPEECH 2023},
  pages={2143--2147},
  doi={10.21437/Interspeech.2023-1152}
}

Online: ISCA Archive

@techreport{ward2023a,
  title = {Dialogs {{Re-enacted Across Languages}}, {{Version}} 2},
  author = {Ward, Nigel G. and Avila, Jonathan E. and Rivas, Emilia and Marco, Divette},
  year = {2023},
  number = {UTEP-CS-23-27},
  institution = {{University of Texas at El Paso}}
}

Online: UTEP, Nigel Ward

Included is a Flake8 Python linter settings file (.flake8). If VS Code is your preferred editor, included is a VS Code workspace settings file (.vscode) with debug configurations. I suggest installing the markdownlint extension for linting Markdown, and formatting Python code on save with Black.