Software supporting "Modeling Vocal Entrainment in Conversational Speech using Deep Unsupervised Learning"
original code by Md Nasir Modified for Python3 by Megh Krishnaswamy and Adarsh Pyrelal
- Bash
- OpenSMILE
- sph2pipe
- Kaldi
- LDC corpus and metadata files
(for ToMCAT users, a small sample of the corpus is available atkraken.sista.arizona.edu:/media/mule/projects/ldc
)
note: move all files in the
- feat/ :- Directory containing scripts for acoustic feature extraction and functional computation
- models/ :- Directory containing scripts for different deep unsupervised learning models for entrainment
- utils/ :- Directory containing utility files used by other scripts
ToDo- edit this with latest files
- ~~
entrainment_config.py
- ~~
feats/run_all_nopre.sh
- ~~
models/NED/run_all_nopre.sh
ToDo- edit this for instructions for the makefile
- Setup and activate virtual environment
- Run
pip install -e .
- To run the code on your system, download and set-up the LDC data, and access/create
Fisher_meta.csv
scp -r [username]@kraken.sista.arizona.edu:/media/mule/projects/ldc [local directory]
- Add a step for setup.py, wheel and installing requirements
- ~~Ensure you have installed all required python libraries
- ~~Edit
entrainment_config.py
,feats/run_all_nopre.sh
andmodels/NED/run_all_nopre.sh
with relevant filepaths - ~~Run
feats/run_all_nopre.sh
andmodels/NED/run_all_nopre.sh
to extract OpenSMILE features for all sound files - Run
models/triplet/prep/create_kaldi_files.py
to extract Kaldi files - Open feats directory to follow the steps for modelling
ToDo- edit this to reflect the new files Make sure the following directories/files have permissions:
- chmod 777 feats
- chmod 777 model/NED
- chmod 777 feats/emobase2010_haoqi_revised.conf
- chmod 777 models/NED/emobase2010_revised.conf ~~3. chmod 755 feats/run_all_nopre.sh ~4. chmod 755 models/NED/run_all_nopre.sh