This repository contains supplementary material for my presentation “Automatische Optimierungsverfahren in der akustischen Analyse soziophonetischer Feldaufnahmen” at the (virtual) P&P 2020, 16. Tagung zur Phonetik und Phonologie im deutschsprachigen Raum, Trier. <www.pundp2020.uni-trier.de>
Daniel Duran
duran@leibniz-zas.de https://orcid.org/0000-0001-6769-2281
- internals: reimplemented the call of Nelder_Mead and the handling of previously evaluated parameter combinations
- a working installation of Praat http://www.fon.hum.uva.nl/praat/
- a working installation of R https://www.r-project.org/
- the R package foreach https://CRAN.R-project.org/package=foreach
- the R package doMC (on Linux) https://CRAN.R-project.org/package=doMC
- the R package doSNOW (on Windows) https://CRAN.R-project.org/package=doSNOW
- the R package lme4 https://cran.r-project.org/package=lme4
- the R package textgRid https://cran.r-project.org/package=textgRid
- R packages from tidyverse https://cran.r-project.org/package=tidyverse
- audio (wav) and TextGrid files (not provided)
Semi-supervised optimization of Praat analysis parameters for voicing measurements by recording room.
In order to run the optimization, you only need to run the R-script optVoicing.R.
WARNING: Running this script may take a very long time (depending on your machine). You may want to run the script first on a smaller subset of your data and/or restrict the maximum number of function evaluations to a very small value until your setup is ready!
The main configuration is done within the R-script, by setting the respective parameters in the list-object KONFIG at the beginning of the script. Specifically, you need to set the paths to the Praat executable, the path to the Praat script voice-Advanced.praat, and the path to the input files table CSV-file: EXAMPLEvoicingFiles.csv.
If the script crashes or if it is aborted for some reason, you may chose to initialize the search with previously found optimal parameters.
- The number of available voice labels per room is not checked. This implementation requires a minimum of 2 tokens per label (in function load_gold_voice_annotations). However, a (much) larger number should be provided for reasonable optimization results.
- The train/test split is random. Set seed to achive identical results in different runs.
- Praat runs always on all data. The train/test split is taken into accout only for the evaluation.
Unsupervised optimization of Praat analysis parameters for formant measurements by speaker.
In order to run the optimization, you only need to run the R-script optFormants.R.
WARNING: Running this script may take a very long time (depending on your machine). You may want to run the script first on a smaller supset of your data until your setup is ready!
The main configuration is done within the R-script, by setting the respective parameters in the list-object KONFIG at the beginning of the script. Specifically, you need to set the paths to the Praat executable, the path to the Praat script extractFormantsChunked.praat, and the path to the input files table CSV-file: EXAMPLEformantFiles.csv.
This work was funded by the German Research Foundation (DFG), within the research project Soziophonetische Untersuchungen zum deutschen Multi-Ethnolekt; grant to Prof. Dr. Dr. h.c. Peter Auer.
- Auer, Peter, Daniel Duran, and Christina Davril: “Voicing and Vowel Quality in the German (Multi-)Ethnolect.” Presentation presented at the International Conference on Language Variation in Europe (ICLaVE 10), Leeuwarden (NL), June 2019. https://www.fryske-akademy.nl/en/iclave100
- Douglas Bates, Martin Mächler, Ben Bolker and Steve Walker: “Fitting Linear Mixed-Effects Models Using lme4.” Journal of Statistical Software, 67:1, 2015: 1-48. DOI: 10.18637/jss.v067.i01
- Boersma, Paul, and David Weenink: “Praat: Doing Phonetics by Computer.” http://www.praat.org
- Kisler, Thomas, Uwe Reichel, and Florian Schiel: “Multilingual Processing of Speech via Web Services.” Computer Speech & Language 45 (September 2017): 326–47. DOI: 10.1016/j.csl.2017.01.005
- Microsoft and Steve Weston: “foreach: Provides Foreach Looping Construct.” R package version 1.4.7. 2019. https://CRAN.R-project.org/package=foreach
- Nelder, J. A., and R. Mead. “A Simplex Method for Function Minimization.” The Computer Journal, 7:4, 1965: 308–13. DOI: 10.1093/comjnl/7.4.308
- Hong Ooi, Microsoft Corporation and Stephen Weston: “doSNOW: Foreach Parallel Adaptor for the 'snow' Package.” R package. 2019. https://CRAN.R-project.org/package=doSNOW
- Patrick Reidy: “textgRid: Praat TextGrid Objects in R.” 2016. R package. https://CRAN.R-project.org/package=textgRid
- R Core Team: “R: A Language and Environment for Statistical Computing.” R Foundation for Statistical Computing. Vienna, Austria. 2019. https://www.R-project.org
- Revolution Analytics and Steve Weston: “doMC: Foreach Parallel Adaptor for 'parallel'.” R package version 1.3.6. 2019. https://CRAN.R-project.org/package=doMC
- Schiel, Florian: “Automatic Phonetic Transcription of Non-Prompted Speech.” In 14th International Congress of Phonetic Sciences (ICPhS-14), 607–10, 1999. https://www.internationalphoneticassociation.org/icphs-proceedings/ICPhS1999/p14_0607.html
- Hadley Wickham et al.: “Welcome to the tidyverse.” Journal of Open Source Software, 4:43, 2019: 1686. DOI: 10.21105/joss.01686