CLARIN-NL - Phonetics

TQE

root — Mon, 08 Apr 2019 08:38:55 +0000

Title

TQE: Transcription Quality Evaluation

Description

The Transcription Quality Evaluation (TQE) tool is an instrument that automatically evaluates the quality of phonetic transcriptions. The application makes it possible to upload pairs of files consisting of an audio file and a transcription file and process them as follows: the audio signal and the phonetic transcription are aligned, segment boundaries are derived for each phone, and for each segment-phone combination it is determined how well they fit together, i.e. for each phone a TQE measure (a confidence measure) is determined, a number ranging from 0-100%, indicating how good the fit is, i.e. the quality of the phone transcription. The higher the number, the better the fit. The output of the TQE tool consists of a TQE measure and the segment boundaries for each phone in the corpus. The TQE tool thus makes it possible to find (sequences of) segments for which the match of the phone symbols with the audio signal is not optimal, in other words, the TQE tool can be used to check the quality of phonetic transcriptions. This can be useful for validating (manual) phonetic transcriptions, but also to compare and select (‘competing’) transcriptions, e.g. to study pronunciation variation. The TQE tool can thus be usefully applied in all research – in various (sub-) fields of humanities and language and speech technology (L&ST) – in which audio and phonetic transcriptions are involved.

Project

TQE

CLARIN National Project

CLARIN-NL

CLARIN centre

MPI for Psycholinguistics

Country

Netherlands

Tool Type

web application

Research Phase

Enriching Data

Tool status

published

Input format

audio/wav

text/plain

Output format

text/plain

Input Language

English

Access Contact

Project Contact

Dr. H. Strik

Creator Contact

Dr. H. Strik

Documentation

Manual

Source code

not specified

Original source

http://portal.clarin.nl/node/1912

Resource

CMDI File Link

License

unknown

Inventory Scope

local

Gabmap

root — Mon, 08 Apr 2019 08:38:48 +0000

Title

Description

Gabmap is a free web-based application for dialectometry. It measures the differences in sets of phonetic (or phonemic) transcriptions via edit distance. Gabmap has a graphical user interface that makes string comparison facility available as a web application. This enables wider experimentation with the techniques. Gabmap (a.k.a. ADEPT) measures pronunciation distances based on transcriptions and aligns pronunciation transcription data. Because the measurements are numeric, they can be aggregated in order to obtain an estimation of overall pronunciation differences among varieties. The software uses a range of edit distance (or Levenshtein) algorithms. It is useful for dialectologists, and has been used extensively in dialectology. It has occasionally been used for other purposes, e.g. trying to identify loan words automatically (Paris, Musée de l’Homme, central Asian project involving Turkic and also Indo-Iranian languages). The software has also been used as the basis of a program to multi-align pronunciation data for the purpose of phylogenetic analysis. The Gabmap developers claim that the program could also be used to measure deviant pronunciation e.g. of second-language learners, or of speakers with speech defects. A variety of related algorithms are implemented in the package of C programs (and R programs) the developers turned into a web application, including a basic version regarding segments only as same or different, and other versions variously respecting consonant/vowel distinctions; using phonetic segment distances as provided via an assignment of phonetic or phonological features to segments; using segment distances as learned from refining alignment correspondences; and applying weightings derived from (inverse) frequency (derived from Goebl’s work) or depending on the position within a word. There are useful auxiliary programs aimed at assisting users in converting phonetic data to X-SAMPA and at spotting errors. (In working with users in the past, the developers have noted that data conversion is a major hurdle.) There are additional meta-analytical calculations aimed at gauging how reliable the signal is from a given set of data, and aimed at comparing various options with respect to the degree to which they capture the geographic cohesion one assumes in dialectology. Gabmap was developed in the CLARIN-NL project ADEPT: Assaying Differences via Edit-Distance of Pronunciation Transcriptions.

Project

ADEPT

CLARIN National Project

CLARIN-NL

CLARIN centre

Meertens/HuC

Research domain

Linguistics

Linguistic Subject

Dialectology

Phonetics

Phonology

Tool task

visualisation

analysis

Country

Netherlands

Tool Type

web application

Research Phase

Data analysis

Tool status

published

Output format

application/pdf

Input Language

Language independent

Access Contact

/mailto:info@gabmap.nl/

Meertens Institute

Project Contact

Dr. Charlotte Gooskens

Creator Contact

Peter Kleiweg

Documentation

Gabmap Tutorial

Demonstrator Scenario

Gabmap: Quick Start

Source code

not specified

Original source

http://portal.clarin.nl/node/1917

Publications

Nerbonne, J., Colen, R., Gooskens, C., Kleiweg, P., and Leinonen, T. (2011). Gabmap — A Web Application for Dialectology. Dialectologica, Special issue II, 65-89.

T. Leinonen, Ç. Çöltekin, J. Nerbonne, Using Gabmap. Lingua Vol. 178, 71-83, doi:10.1016/j.lingua.2015.02.004

Resource

CMDI File Link

License

unknown

Inventory Scope

local

D-LUCEA

Erica Renckens — Mon, 19 Jan 2015 16:27:31 +0000

D-LUCEA: Database of the Longitudinal Utrecht Collection of English Accents

Summary

The LUCEA database is a database of existing speech recordings of L1 and L2 speakers of English. The recorded speakers are students from an international student community where English is used as lingua franca. These students are being recorded longitudinally throughout their 3-year period on campus, using read and spontaneous speech in L1 and in L2 English (or in L1 English only). The database is of interest for research and development in linguistics, language education (pronunciation training), speech technology (foreign accent detection, language recognition, speech recognition), and sociophonetics.

Background

The LUCEA Database consists of speech recordings of students at the University College Utrecht (UCU), the international honours college of Utrecht University, the Netherlands. At UCU, English is the language of instruction and also the social lingua franca. So far, 72 (cohort 2010) and about 70 (cohort 2011) students have been recruited, as well as 17 exchange students who are native speakers of English. All speaker groups are about 70% female and 30% male. For the 2010 students, 3 recordings are available. For the 2011 students and for the exchange students, 1 recording is available.

Each recording consists of approx 20 minutes of speech, half of which is read (English read texts and sentences, and L1 read text for 2011 cohort) and half of which is spontaneous (2 minutes English monologue on informal topic, 2m monologue in L1 on informal topic, 2m English monologue on formal topic, 3m English conversation with interviewer). Each recording is made using a strict protocol describing the exact location of microphones, speakers’ instructions, recording procedure, etc.
The speech recordings are not yet annotated nor transcribed. The transcripts of the read texts are available in PDF. The metadata scheme developed in this project allows future additions of annotations.

The recordings are part of the larger and ongoing LUCEA project in which the recruited students are being followed throughout their 3-year period on the UCU campus. This will result in about 900 recording sessions. At present, about 280 [(72+65+55) + (~70) + (17)] recordings are available.

Contacts

Project leader: Dr Hugo Quené (Utrecht University)  
CLARIN center: Max Planck Institute for Psycholinguistics, Nijmegen

Links

Web-sites: http://lucea.wp.hum.uu.nl/
User scenario's (screencasts, screenshots): n.a.
Manual: n.a.
Tool/Service link: soon on http://corpus1.mpi.nl/
Data Link (VLO): http://catalog.clarin.eu/vlo/search?q=D-LUCEA
Publications:
- Orr, R and Quené, H. 2017. D-LUCEA: Curation of the UCU Accent Project Data. In:Odijk, J and van Hessen, A. (eds.) CLARIN in the Low Countries, Pp. 181–193. London: Ubiquity Press. DOI: https://doi.org/10.5334/bbi.15. License: CC-BY 4.0
- Orr, R., H. Quene, R. Beek, T. Diefenbach, D. van Leeuwen, M. Huijbregts (2011), An international English speech corpus for longitudinal study of accent development, Proceedings Interspeech 2011, pp. 1889-1892.
- Orr, R., Huijbregts, M., Beek, R. van, Teunissen, L., Backhouse, K., Leeuwen, D. van (2014), Semi-automatic annotation of the UCU accents speech corpus. LREC 2014: 1483-1487.
- Quené, H. & Orr, R. (2014), Long-term convergence of speech rhythm in L1 and L2 English. Speech Prosody 2014.

Country

Netherlands

CLARIN centre

MPI for Psycholinguistics

Language

English

Research domain

Linguistics

Phonetics

Acquisition

Resource tags

data

corpus

multi-media/muliti-modal

mono-lingual

Annotations

Linguistic Annotation

Orthographic

Orthographic Transcription

Format

Textual data

CSV

XML

Media data

wav

Gabmap

root — Sun, 20 Apr 2014 15:26:35 +0000

ADEPT: Assaying Differences via Edit-Distance of Pronunciation Transcriptions - Gabmap

Summary

Background

Gabmap (a.k.a. ADEPT) measures pronunciation distances based on transcriptions and aligns pronunciation transcription data. Because the measurements are numeric, they can be aggregated in order to obtain an estimation of overall pronunciation differences among varieties. The software uses a range of edit distance (or Levenshtein) algorithms. It is useful for dialectologists, and has been used extensively in dialectology. It has occasionally been used for other purposes, e.g. trying to identify loan words automatically (Paris, Musée de l’Homme, central Asian project involving Turkic and also Indo-Iranian languages). The software has also been used as the basis of a program to multi-align pronunciation data for the purpose of phylogenetic analysis. The Gabmap developers claim that the program could also be used to measure deviant pronunciation e.g. of second-language learners, or of speakers with speech defects.
A variety of related algorithms are implemented in the package of C programs (and R programs) the developers turned into a web application, including a basic version regarding segments only as same or different, and other versions variously respecting consonant/vowel distinctions; using phonetic segment distances as provided via an assignment of phonetic or phonological features to segments; using segment distances as learned from refining alignment correspondences; and applying weightings derived from (inverse) frequency (derived from Goebl’s work) or depending on the position within a word. There are useful auxiliary programs aimed at assisting users in converting phonetic data to X-SAMPA and at spotting errors. (In working with users in the past, the developers have noted that data conversion is a major hurdle.) There are additional meta-analytical calculations aimed at gauging how reliable the signal is from a given set of data, and aimed at comparing various options with respect to the degree to which they capture the geographic cohesion one assumes in dialectology.

Contacts

Project leader:  Dr. Charlotte Gooskens (RUG)
CLARIN center: Meertens Institute
Help contact : info@gabmap.nl (General info), support@gabmap.nl (Technical support)

Links

Web-sites: http://www.gabmap.nl/
User scenario's (screencasts, screenshots): http://www.gabmap.nl/~app/doc/IntroVideo/
Manual: http://www.gabmap.nl/~app/doc/tutorials/tutorial/index.htm, http://www.gabmap.nl/~app/doc/preparing/ (quick start)
Tool/Service link: http://www.gabmap.nl/~app/bin/home
Publications: Nerbonne, J., Colen, R., Gooskens, C., Kleiweg, P., and Leinonen, T. (2011). Gabmap — A Web Application for Dialectology. Dialectologica, Special issue II, 65-89.

Research domain

Linguistics

Phonetics

Phonology

Dialectology

Language

Language independent

Resource tags

service

web-application

Tool task

analysis

visualisation

CLARIN centre

Meertens/HuC

Country

Netherlands

TQE

root — Sun, 20 Apr 2014 13:45:49 +0000

TQE: Transcription Quality Evaluation

Summary

The Transcription Quality Evaluation tool can be used to check the quality of phonetic transcription. The only thing the researcher has to do is upload pairs of files consisting of an audio file and a transcription file. After uploading he’ll receive an e-mail with the matching output.

Background

The Transcription Quality Evaluation (TQE) tool is an instrument that automatically evaluates the quality of phonetic transcriptions. The application makes it possible to upload pairs of files consisting of an audio file and a transcription file and process them as follows: the audio signal and the phonetic transcription are aligned, segment boundaries are derived for each phone, and for each segment-phone combination it is determined how well they fit together, i.e. for each phone a TQE measure (a confidence measure) is determined, a number ranging from 0-100%, indicating how good the fit is, i.e. the quality of the phone transcription. The higher the number, the better the fit. The output of the TQE tool consists of a TQE measure and the segment boundaries for each phone in the corpus.
The TQE tool thus makes it possible to find (sequences of) segments for which the match of the phone symbols with the audio signal is not optimal, in other words, the TQE tool can be used to check the quality of phonetic transcriptions. This can be useful for validating (manual) phonetic transcriptions, but also to compare and select (‘competing’) transcriptions, e.g. to study pronunciation variation. The TQE tool can thus be usefully applied in all research – in various (sub-) fields of humanities and language and speech technology (L&ST) – in which audio and phonetic transcriptions are involved.

Contacts