Date: October 29, 2025 (Wed)
Location: DH-203/204, International Christian University
Language: English (Japanese ok)
Registration Fee: Free
日付：2025年10月29日（水）
場所：国際基督教大学ダイアログハウス２０３・２０４
言語：英語（日本語可）
登録費：無料
Montreal Forced Aligner (MFA) Workshop:
Tutorial and Practice
10:30-12:30 (part 1, hybrid)
- Lecture title: Speech-to-text alignment with the Montreal Forced Aligner
Abstract: Speech-to-text alignment is the first step in many phonetic analysis pipelines. Montreal Forced Aligner provides timestamps of words and speech sounds from audio and verbatim transcripts for 20+ pretrained languages, along with model adaptation and training functionality. In this talk, I will cover basic MFA usage and commands to fix commonly encountered issues when applying MFA to new data.
14:00-17:00 (part 2, in-person only)
- Bring your own data (if you don't have data set, we will provide some)
- Workshop title: One size does not fit all: Tailoring MFA to your data
Abstract: This workshop will focus on practical usage of applying MFA to new data. The first 30 minutes will be dedicated to making sure MFA is correctly installed and runnable. The remaining time will be focused on running through tutorials for generating alignments and adapting pretrained models to new languages. Demo corpora for English, Mandarin, and Japanese are available to be used, but I recommend bringing your own data to the workshop and attempting to get it aligned, even if it’s not a language supported by MFA out of the box.
