This is a follow-up on a previous post about experiments using automatic transcription for Yongning Na. It’s been an eventful and exciting time!
Oliver Adams has now released his automatic phoneme transcription tool. It’s all open-source code. Here is the address: https://github.com/oadams/persephone
The transcription tool is named Persephone (/pərˈsɛfəni/), after the goddess who was abducted by Hades and must spend one half of each year in the Underworld. Which of linguistics or computer science is Hell, and which the joyful world of Spring and light? For each it’s the other, of course. The name Persephone is intended as a tongue-in-cheek allusion to the sense of estrangement and uncanniness that one can occasionally feel in interdisciplinary collaboration. Interdisciplinary work has difficulties and rewards of its own – like fieldwork, in a way, but to some fieldworkers the difficulties can be perceived as the last straw, coming on top of a heavy work load. It is clear that linguists engaged in language documentation cannot spend half their lives learning about language processing, any more than computer scientists would want to spend half their lives reading books of linguistics. But linguists can be convinced to spend… maybe one fifth of their time playing with software and understanding how state-of-the-art tools work. “Digital Humanities drive the scholarly tradition towards a more uniform approach or an approach where the divide between two cultures – humanities on the one side, and computing on the other – is no longer constructive”, as a report of the Research Data Alliance puts it.
There is a presentation and user guide written by Oliver on the project homepage on GitHub. A paper (LREC 2018) is available here.