A semi-automatic workflow for orthographic transcription and syllabic segmentation

Francesco Cangemi; Jessica Fründt; Harriet Hanekamp; Martine Grice

doi:10.17469/O2106AISV000024

Autori

Francesco Cangemi IfL Phonetics, University of Cologne, Germany https://orcid.org/0000-0003-1016-5178
Jessica Fründt IfL Phonetics, University of Cologne, Germany
Harriet Hanekamp IfL Phonetics, University of Cologne, Germany
Martine Grice IfL Phonetics, University of Cologne, Germany https://orcid.org/0000-0003-4973-4059

DOI:

https://doi.org/10.17469/O2106AISV000024

Parole chiave:

orthographic transcription, forced alignment, syllabic segmentation, spontaneous speech, automation

Abstract

Manual orthographic transcription of spontaneous speech is notoriously time consuming, and segmentation at the level of the syllable requires further large amounts of processing time. Automatic orthographic transcription and automatic syllable segmentation, on the other hand, usually yield unsatisfactory precision, especially when applied to spontaneous speech. In this contribution, we report on a semi-automatic workflow that uses freely available software and combines the speed of automatic processing with the quality of manual transcription and segmentation. We apply the procedure to highly spontaneous speech and show (i) no loss of quality compared with manual output and (ii) an average 75% reduction of manual processing time.