The code of AcousticGender.space is hydra-hosted, meaning that it is available on multiple code hosting platforms, with none held above the others as the canonical source. We take advantage of the decentralized design of Git so that if one host goes down or stops supporting us, another will always be available.
The repository links are temporarily disabled to comply with anonymization requirements of our paper submission.
While our code is continuously being improved, for our research to be replicable, it is important that the code and data be available to researchers in the exact form it was in used in experiments. Therefore, we offer zip archives frozen in that state:
code.zip (32M) - The Python code used to generate process recordings and generate the formant weights used in the study, as well as the resources and code of the web frontend.
corpus.zip (105M) -
A selection of recordings from the GMU Speech Accent Archive.
The values in stats.json were obtained
by running corpus-analysis.py
on the contents of this archive.
corpus-processed.zip (540M) -
The above corpus after being processed by corpus-analysis.py.
Includes the recordings after undergoing noise removal and reformatting
-- as well as the formant frequency data and textgrids
aligning the recordings to transcripts.
ex1.zip (3.3M) - The Python code used to process recordings in the study as well as the resources and code of the web frontend.