Accueil > Evénements


Evénement No. 1584
http://www.lpl-aix.fr/event/1584

Annotations : méthodes, outils et interprétation

Type :

séminaire LPL

Auteur :

Katarzyna *Klessa *(1)
Imed *Laaridh* (2)
Brigitte *Bigi *(3)

Affiliation :

(1) Institute of Linguistics, Department of Phonetics, Adam Mickiewicz University of Poznan, Poland
(2) Laboratoire Informatique d'Avignon, Brain and Language Research Institute, Avignon, France
(3) Laboratoire Parole et Langage, Aix-en-Provence, France

Lieu ou publication :

salle B011, LPL

Pays :

France

Date (de début) :

6/2/2015

Date de fin :

Heure :

URL:

Conditions d'accès :

 Organisateur(s) :

Brigitte Bigi

 Partenaires :

 Email pour contact :

Présentation :

Présentation générale :

Dans les dernières décennies, l’informatique a transformé l'accès aux données et affecté la démarche d'étude linguistique. C'est désormais un moyen indispensable pour :
- étudier la langue et ses propriétés grâce à l’exploitation de corpus de grande ampleur ;
- structurer et normaliser les connaissances linguistiques (phonétiques, morphologiques, lexicales, syntaxiques, sémantiques, etc.) ;
- valoriser, partager et mutualiser les résultats de la recherche sur une langue, qui passent le plus souvent par la production de ressources et d’outils informatiques.

Les méthodes, mises en pratique par les outils computationnels, délivrent ainsi des résultats, des constatations reproductibles et inattaquables. Mais en quoi ces résultats sont-ils révélateurs ? Comment les interpréter ? Ce séminaire vise à ouvrir le débat sur les solutions pour annoter des corpus, les représenter et les traiter, en allant au delà des solutions classiquement proposées.


Katarzyna Klessa
This presentation describes selected features and example uses of Annotation Pro – a freely available annotation tool (download at: http://annotationpro.org/downloads/). The software extends the potential of a typical language-independent, multi-layer annotation system with a novel component based on a graphical representation of feature space for annotation of continuous and non-categorial features. Among others,this functionality supports annotation of emotions in speech or other features difficult to define in an unambiguous way. In order to enable flexible extensions of the program's functionality (e.g., by automatising the process of annotation or annotation mining), a plugin architecture was developed (for C# plugins). In the present seminar several plugins will be briefly demonstrated, e.g., TGA (Time Group Analysis by Dafydd Gibbon) or SRMA (Segment Rate Moving Average). Both of these extensions are useful for the analysis of timing and speaking rate variability in conversational speech. Apart from multilayer annotation options, the program provides a perception experiment framework aimed especially (but not exclusively) at testing hypotheses related to perception of continuous and non-categorial features in speech.
Apart from providing options for annotation and analysis of recordings such as configurable sound signal display (waveform, several types of spectrograms), the programme also makes it possible to annotate written text corpora (without any speech signal associated). For the sake of interoperability, Annotation Pro exports/imports data to/from several external formats, e.g., Praat's .TextGrid, Transcriber's .TRS, as well as .CSV or .TXT file formats. Both individual files and file collections can be easily exported to spreadsheets or other formats.

Katarzyna Klessa’s bio
Personal website: http://katarzyna.klessa.pl/
Selected publications: http://katarzyna.klessa.pl/?page_id=33
Brief CV: http://katarzyna.klessa.pl/?page_id=4/

Imed Laaridh
Perceptual evaluation is still the most common method in clinical practice for diagnosing and following the progression of the condition of patients with speech disorders. Automatic speech processing-based tools are largely used in the literature to deal with dysarthric speech. Mainly, the goal of such approaches is to provide patients withassistive technologies or to provide technologies for an objective assessment of the dysarthric speech severity. In this presentation, an original approach to detect automatically abnormal acoustic zones and phenomena in speech signal produced by patients suffering from speech disorders is proposed. This automatic process aims at enhancing the manual investigation of human experts while at the same time reducing the extent of their intervention, by calling their attention to specific parts of the speech, considered as atypical from an acoustical point of view.

Brigitte Bigi
Annotating corpora is of crucial importance in Corpus Linguistics. More and more annotated corpora are now available, and so are tools to annotate automatically and/or manually. However, hard-to-annotate phenomena imply uncertainty while annotating: imprecision or indeterminacy are intrinsic to the annotation process. In this presentation, a general description level representing the multimodal linguistic annotations will be presented. It focuses on time and data content representation: This presentation reconsiders and enhances the current and generalized representation of an annotation. This proposal is implemented in SPPAS.

Programme :

Programme :

9h15 – Accueil des participants

9h25 – Présentation du séminaire

9h30 – Katarzyna Klessa (AMU, Poznan)

10h20 – Pause café

10h35 – Imed Laaridh (LIA-BLRI, Avignon)

11h10 – Brigitte Bigi (LPL, Aix-en-Provence)

11h50 – Conclusion du séminaire

12h00 – Apéritif offert


 Remarques :

De 14h00 à 16h00, Katarzyna Klessa sera disponible pour présenter le logiciel AnnotationPro (voir http://annotationpro.org/) et en discuter.

A cette occasion, vous pouvez venir avec vos propres données annotées !


Nom de l'auteur de la fiche :

Caterina PETRONE

[Modifier cette fiche] [Créer un nouvel enregistrement]