Automatic phonetic labelling & segmentation of multiple Hungarian recordings
Interviews and conversation are often recorded and later transcribed. The web service WebMAUS Multiple available in the CLARIN infrastructure provides tools to combine audio recordings and transcriptions in a way that the words and audio signals are time aligned. In contrast to WebMAUS Basic this services allows a multitude of options and more languages, and also provides a batch processing mechanism to segment whole speech corpora.
Especially relevant for
- Speech technologists
- Media science
50 recordings of Hungarian (*.wav) and 50 orthographic transcripts (*.txt)
50 hierarchical segmentations (words, phonology, phones) in praat compatible format (*.TextGrid); phonetic segments should be encoded in IPA
Web Interface BAS WebMAUS Multiple
Related CLARIN-D tools and services
Short guide on how to use BAS WebMAUS Multiple
- download the ZIP package ftp://clarin.phonetik.uni-muenchen.de/BASWebServices/useCases/multipleFileSegmentation.zip and
- un-pack it onto your local desktop folder; there should be a directory called 'multipleFileSegmentation' on your desktop.
By Web Interface:
- go to http://clarin.phonetik.uni-muenchen.de/BASWebServices
- select service 'WebMAUS Multiple'
- open the directory 'multipleFileSegmentation/Corpus' on your desktop; select all 50 *.wav and 50 *.txt files; drag&drop them to the designated drop area of the service. The drop might take a few seconds, so please be patient. All 100 files should be visible in the drop area after completion.
- press button 'Upload' after the upload is completed you can inspect the uploaded text files or listen to the uploaded signals by clicking on the file link in the drop area.
- execute WebMAUS with the following options (no need to adjust the other options)
- Language = Hungarian
- KAN tier in TextGrid = true
- ORT tier in TextGrid = true
- Chunk segmentation = force
- Output symbols = ipa
- Confirm the terms-of-usage and press the button 'Run Web Service'
- after completion the resulting 50 *.TextGrid files are shown below; click on "Download as ZIP-File" to download all files to your local system
- start praat (download at praat.org) on your local system and open *.TextGrid *.wav pairs to inspect the segmentation