iiinova — where your digital self keeps evolving

Michael McAuliffe

@memcauliffe.com

206 followers 340 following 31 posts

Linguist, computational phonetician, recovering academic. Maintainer of Montreal Forced Alignment Sometimes my tinkerings are useful memcauliffe.com github.com/mmcauliffe he/him

Posts Replies Media

Michael McAuliffe @memcauliffe.com · 10/03/2024

@carlbergstrom.com pardon for the ping, but is this a corvid of some kind? I don't recognize it and I'm not having luck tracking it down on the Internet. This is in South Florida.

110

Michael McAuliffe @memcauliffe.com · 30/09/2023

Phew, finally got a basic pipeline that I'm happy with for a new Japanese MFA model. Uses sudachipy for morphological parsing with postprocessing to create phonological words. A G2P model and phonological rules create pronunciations from the katakana pronunciations from sudachipy.

Force aligned audio file for Japanese from recombining morphemes into phonological words and generating pronunciations as necessary with a G2P model and lexical phonological rules

030

Michael McAuliffe @memcauliffe.com · 03/09/2023

Getting some good progress on diarizing the Japanese LaboroTV corpus since it doesn't have any speaker metadata. Maybe not the most efficient use of my time, but the interface for using SpeechBrain speaker recognition models and PLDA from kalpy is coming along!

Anchor annotator tool's interface for sorting utterances by speaker

060