Sujet de stage / thèse : Apprentissage dynamique de caractéristiques par champs aléatoires conditionnels pour la détection des sentiments dans les conversations téléphoniques

ENCADREMENT : Chloé Clavel[1] , Slim Essid [2]

LIEU: Telecom ParisTech, 37 rue Dareau, 75014 Paris

Date de début du stage : début 2015

Financement de thèse de doctorat garanti

ÉQUIPE(S) D?ACCUEIL DE LA THÈSE :

département TSI, équipe MultiMedia (MM),

département TSI, équipe Audio Acoustique et Ondes (AAO)

MOTS CLÉS: analyse des sentiments, opinion mining, machine learning, champs aléatoires conditionnels, traitement automatique des langues, traitement de la parole

SUJET:

Le domaine du sentiment analysis et de l?opinion mining  est un domaine en plein essor avec l?arrivée en masse de données textuelles sur le web comportant des expressions d?opinions par les citoyens (critiques de films, débats sur les commentaires de forums, tweets) [14]. Les recherches en traitement automatique des langues se mobilisent sur le développement de méthodes de détection d?opinions dans les textes en s?appuyant sur ces nouvelles ressources. Cependant, le problème de la détection des opinions et des sentiments se limite souvent a? un proble?me de classification sur les axes de valence (positif, ne?gatif) et d?intensite? [11]. De plus, l?analyse des sentiments dans la parole est un proble?me jusqu?ici peu e?tudie? et qui ne?cessite la prise en compte non seulement des spe?cificite?s langagie?res de l?oral et des erreurs des syste?mes de reconnaissance mais aussi l?inte?gration des caracte?ristiques prosodiques dans le contexte verbal. Enfin, les méthodes classiquement développées dans le domaine de la fouille d?opinions ne s?attaquent pas a? la question du temps de de?tection, question essentielle dans le cadre de l?analyse des centres d?appels qui requiert non seulement une détection en ligne pour lever des alertes mais aussi une fouille des données pour une meilleure connaissance client.

Le sujet de stage porte sur le développement de méthodes d'analyse des sentiments dans des conversations téléphoniques. Le stagiaire pourra s?appuyer sur la the?orie de l?appraisal adapte?e au contenu verbal telle que de?finie par les psycho-linguistes dans [8] pour construire les mode?les computationnels de de?tection d?expressions e?valuatives. Pour ces derniers, des me?thodes de mode?lisation hybrides (apprentissage de caractéristiques piloté par des connaissances a priori) a? base de CRF (Conditional Random Fields ou champs ale?atoires conditionnels) pourront être développées. Ces mode?les sont couramment utilise?s dans le domaine du Traitement Automatique du Langage pour l?analyse morpho-syntaxique ou pour la de?tection d?entite?s nomme?es mais sont encore tre?s rarement et partiellement utilise?s dans le contexte de l?analyse des sentiments. Ils sont adapte?s a? la pre?diction de donne?es structure?es et sont de nature discriminative ce qui leur procure une capacite? a? exploiter un tre?s grand nombre de fonctions d?observation (feature functions). Une des originalite?s du stage consistera ainsi a? repre?senter par des fonctions d?observation les règles sémantiques habituellement utilise?es pour les tâches qui nous concernent afin de les valider par un apprentissage a? partir des donne?es. Cet apprentissage permettra a? la fois de de?terminer la façon optimale de combiner les re?gles conside?re?es (notamment par l?emploi de noyaux ou kernels), et de re?aliser une se?lection automatique des plus pertinentes (en munissant les CRF de me?canismes de se?lection de caracte?ristiques).

REFERENCES :

[1] C. Clavel, G. Adda, F. Cailliau, M. Garnier-Rizet, A. Cavet, G. Chapuis, S. Courcinous,

C. Danesi, A. Daquo, M. Deldossi, et al. Spontaneous speech and opinion detection: mining call-centre transcripts. Language Resources and Evaluation, pages 1?37, 2012.

[2] C. Clavel, C. Pelachaud, and M. Ochs. User?s sentiment analysis in face-to-face humanagent interactions - prospects. In Workshop on Affective Social Signal Computing, Satellite of Interspeech, Grenoble, August 2013, 2013.

[3] C. Clavel, I. Vasilescu, L. Devillers, G. Richard, and T. Ehrette. Fear-type emotion recognition for future audio-based surveillance systems. Speech Communication, 50(6):487?503, 2008.

[4] C. Joder, S. Essid, and G. Richard. A conditional random field viewpoint of symbolic audiotoscore matching. In ACM Multimedia 2010, Florence, Italy, October 2010.

[5] C. Joder, S. Essid, and G. Richard. A conditional random field framework for robust and scalable audio-to-score matching. IEEE Transactions on Audio, Speech and Language Processing, 19(8):2385 ? 2397, November 2011.

[6] C. Joder, S. Essid, and G. Richard. Learning optimal features for polyphonic audio-to-score alignment. IEEE Transactions on Audio, Speech, and Language Processing, 21(10):2118?2128, 2013.

[7] R. Lavalley, C. Clavel, and P. Bellot. Extraction probabiliste de chaînes de mots relatives à une opinion. Traitement Automatique des Langues, 2010.

[8] J.R. Martin and P.R.R. White. The language of evaluation. Palgrave Macmillan Basingstoke and New York, 2005.

[9] G. McKeown and M. Valstar. The SEMAINE Database: Annotated Multimodal Records of Emotionally Colored Conversations between a Person and a Limited Agent. IEEE Transactions on Affective Computing, 3(1):5?17, 2012.

[10] Vinciarelli, A., Pantic, M., & Bourlard, H. (2009). Social signal processing: Survey of an emerging domain. Image and Vision Computing, (November 2008).

[11] Tsytsarau, M., & Palpanas, T. (2012). Survey on mining subjective data on the web. Data Mining and Knowledge Discovery.

[12] M. Ochs, Y. Ding, N. Fourati, M. Chollet, B. Ravenet, F. Pecune, N. Glas, K. Prépin, C. Clavel et C. Pelachaud, Vers des Agents Conversationnels Animés Socio-Affectifs, Interaction Humain-Machine (IHM'13), November 2013, Bordeaux, France

[13] Magalie Ochs et Catherine Pelachaud, Socially Aware Virtual Characters: The Social Signal of Smiles, IEEE Signal Processing Magazine, Vol 30 (2), p. 128-132, March 2013

[14] El-Bèze, M., Jackiewicz, A., & Hunston, S. (2010). Opinions, sentiments et jugements d?évaluation. Traitement Automatique Des Langues, 51(3)

PROFIL DU CANDIDAT: étudiant titulaire d?un master 2 recherche

-        Apprentissage statistique / reconnaissance des formes

-        Traitement de la parole, traitement du langage naturel

-        Bon niveau en programmation (Java, C/C++, Python)

-        Bon niveau d?anglais

CANDIDATURES :

à envoyer à chloe.clavel@telecom-paristech.fr, slim.essid@telecom-paristech.fr,:

-        Curriculum Vitae

-        Lettre de motivation personnalisée expliquant l?intérêt du candidat sur le sujet (directement dans le corps du mail)

-        Relevés de notes des années précédentes

-        Contact d?une personne de référence

Les candidatures incomplètes ne seront pas examinées.

Back

Top

A funded internship is available at Idiap Research Institute in Switzerland:

Link:

http://www.idiap.ch/webapps/jobs/ors/applicant/position/index.php?PHP_APE_DR_9e581720b5ef40dc7af21c41bac4f4eb={__TO%3D%27detail%27%3B__PK%3D%2710166%27}


Description:

Internship on inference of high level semantics from prosodic features.

The Idiap speech group works on many aspects of speech processing, including
recognition and synthesis for the purpose of speech to speech translation.
Recently, much of our work is in the area of speech prosody. Prosody is often
ignored in speech recognition, but is vital for synthesis. To this end, we are
developing models of prosody that lend themselves to automatic extraction,
translation and synthesis of speech.

Prosody is also a communicative device in itself; it can infer meaning in
words, can convey emotion and affect, and is indicative of personality. It is
the latter of these that defines the general direction of the current project.

The successful candidate will begin with standard models of prosody and attempt
to infer higher level semantics. We intend to compare representations of
prosody for such inference, and investigate which representations are
appropriate for which semantics, hence which semantics can be inferred.
Possible tools include the usual array of speech processing methods such as
Markov models, and machine learning approaches such as neural networks and
support vector machines.

The candidate should have strong programming skills, ideally familiarity with
Python or C++. An analytic background in computer science or engineering is
desirable. All our speech work is multi-lingual; to this end, in addition to
English, the candidate would benefit from knowledge of another European or
Slavic language. Some knowledge of French is an advantage but not a requirement.

The application deadline is March 31st 2015; however, the position will be
filled as soon as possible. The appointment is for 6 months. Interns receive a
'stipend' of 2000 CHF per month.

--
Phil Garner

Back

Top

The research training group DFG GRK 1624 'Frequency effects in language'
at the University of Freiburg, Germany, is advertising for 5
fully-funded PhD positions starting October 1, 2015.

http://frequenz.uni-freiburg.de/bewerbung. 

 
Michael Schäfer

Back

Top

POSTDOC POSITION AT RADBOUD UNIVERSITY NIJMEGEN

Postdoc Speech Researcher

Your Task

Within the FAME! project (Frisian Audio Mining Enterprise) we have a vacancy for a post-doctoral researcher. FAME! is a project in the Creative Industry program of the Netherlands Organisation for Scientific Research. The goal of the project is to develop Automatic Speech Recognition (ASR) for Frisian, for the purpose of disclosing an archive of over 2500 hours of radio broadcasts. The challenge is to make a single system that can deal with both Frisian and Dutch. The candidate will research methods to bootstrap Frisian ASR from other languages, and will investigate the issue of code-switching. A second research area is longitudinal automatic speaker recognition, in order to track speakers over almost half a century of radio broadcasts, for which new approaches for compensation of ageing need to be developed.

The project is carried out in co-operation with several other project partners in The Netherlands, most notably the Fryske Akademy, and we require that the candidate can work well with other (remote) team members, taking a lead when appropriate but also contributing to tasks of other team members.

The research will need to be documented in scientific publications in appropriate journals and relevant conference proceedings, and in Open Source software tools.

Your Profile

We are looking for candidates with the following profile:

• A PhD in Engineering, Computer Science, Machine Learning, Signal Processing, Computational Linguistics or any other relevant discipline.

• Preferably experience in speech technology

• A good publication record

• Fluency in spoken and written English

• Good comprehension capabilities of scientific papers

• Experience in the Linux OS, git version control, scripting in Python and bash, high performance computing in Julia or Num/SciPy, authoring in LaTex, and cluster computing

• Affinity with the open source community

• The drive to work with other project partners, also on location.

Position

• 3 years contract at the Center for Language and Speech Technology (CLST), 0,7 - 1,0 fte

• Scale 11, max. monthly gross salary €4551 at full time appointment

Interested?

Send your motivation letter and CV ultimately 15 March 2015 to:

Dr Henk van den Heuvel: E h.vandenheuvel@let.ru.nl; T: +31 24 3611686 and/or

Prof. dr David van Leeuwen: E d.vanleeuwen@let.ru.nl; T +31 81 888 702

Application

Vacancy number: 23.020.15. Please include with your application a motivation letter (attn. of drs. M.J.M. van Nijnatten), CV and any required attachments. You should upload these documents online, http://www.ru.nl/newstaff/working_at_radboud/details/details-vacature?taal=uk&recid=543119&pad=%2fnewstaff'

Back

Top

Plusieurs postes de Maîtres de Conférences en Informatique sont à pourvoir au sein du laboratoire CRIStAL de Lille (www.cristal.univ-lille.fr) dont un à l'Université Lille 1 (www.univ-lille1.fr)

Les profils recherche s'inscrivent dans celles du groupe thématique 'Data Intelligence' du laboratoire et les thèmes suivants seront particulièrement considérés : 
- Apprentissage par renforcement / imitation
- Décisions séquentielles
- Systèmes de dialogue parlé homme-machine
- Interaction homme-machine, homme-robot ou homme-avatar
- Traitement de la parole et du langage pour l'interaction

La fiche pour le premier poste se trouve ici : 
http://www.univ-lille1.fr/digitalAssets/40/40200_FSES_27_MCF_1151_CRIStAL.pdf

Les autres postes ont des profils recherche similaires et seront publiés le 26 février. 

Contacts :
Philippe Preux (philippe.preux@univ-lille3.fr) (animateur du groupe thématique Data Intelligence)
Olivier Pietquin (olivier.pietquin@univ-lille1.fr)
 

Back

Top

The European Language resources Distribution Agency (ELDA), a company specialized in Human Language Technologies within an international context, acting as the distribution agency of the European Language Resources Association (ELRA), is currently seeking to fill an immediate vacancy for a Language Resources Project Manager ? Junior position. This yields excellent opportunities for young, creative, and motivated candidates wishing to participate actively to the Language Engineering field.

Language Resources Project Manager - Junior

Under the supervision of the Language Resources Sales Manager, the Language Resources Project Manager ? Junior will be in charge of the identification of Language Resources (LRs) and the negotiation of rights in relation with their distribution.

The position includes, but is not limited to, the responsibility of the following tasks:
?    Identification of LRs and Cataloguing
?    Negotiation of distribution rights, including interaction with LR providers, drafting of distribution agreements, definition of prices of language resources to be integrated in the ELRA catalogue
?    LR Packaging and Archiving

Profile:
?    University degree or equivalent in computational linguistics, information science, knowledge management or similar fields
?    Experience in project management (especially European projects), as well as practice in contract and partnership negotiation at an international level, would be a plus
?    Dynamic and communicative, flexible to combine and work on different tasks
?    Ability to work independently and as part of a team, in particular the ability to supervise members of a multidisciplinary team
?    Proficiency in French and English
?    Citizenship of (or residency papers) a European Union country

Applications will be considered until the position is filled. The position is based in Paris.

Salary: Commensurate with qualifications and experience.

Applicants should email a cover letter addressing the points listed above together with a curriculum vitae to:

Valérie Mapelli
ELRA / ELDA
9, rue des Cordelières
75013 Paris
FRANCE
Mail : job@elda.org

ELRA was established in February 1995, with the support of the European Commission, to promote the development and exploitation of Language Resources (LRs). LRs include all data necessary for language engineering, such as monolingual and multilingual lexica, text corpora, speech databases and terminology. The role of this non-profit membership Association is to promote the production of LRs, to collect and to validate them and, foremost, make them available to users. The association also gathers information on market needs and trends.

For further information about ELDA/ELRA, visit http://www.elra.info

Back

Top

Deux postes de Maître de Conférences en informatique (section 27) sont mis au concours
pour la campagne 2015 au Centre d'Enseignement et de Recherche en Informatique de
l'Université d'Avignon (ceri.univ-avignon.fr).

Les personnes recrutées effectueront leur recherche au sein du LIA (lia.univ-avignon.fr).

Les profils des 2 postes sont consultables sur lia.univ-avignon.fr rubrique Emplois et
sur GALAXIE :

- Informatique - Profil langage/systèmes complexes  (Poste 0101 / Galaxie 4062) -
https://www.galaxie.enseignementsup-recherche.gouv.fr/ensup/ListesPostesPublies/ANTEE/2015_1/0840685N/FOPC_0840685N_4062.pdf

- Informatique - Profil recherche opérationnelle et optimisation (Poste 0210 / Galaxie
4063) -
https://www.galaxie.enseignementsup-recherche.gouv.fr/ensup/ListesPostesPublies/ANTEE/2015_1/0840685N/FOPC_0840685N_4063.pdf

Contacts :

Enseignement : Corinne Fredouille corinne.fredouille@univ-avignon.fr - Directrice des
Etudes du CERI et Fabrice Lefèvre fabrice.lefevre@univ-avignon.fr - Directeur du CERI

Recherche : Georges Linares georges.linares@univ-avignon.fr - Directeur du LIA et Fabrice
Lefèvre fabrice.lefevre@univ-avignon.fr - Directeur du CERI

Back

Top

One post-doc grant is now available in project INSIDE: Intelligent Networked Robot Systems for Symbiotic Interaction with Children with Impaired Development. The project is a collaboration between Carnegie Mellon University, INESC-ID, Instituto Superior Técnico, Universidade Nova de Lisboa, Hospital Garcia de Orta, and three Portuguese companies (IDMind, VoiceInteraction and Plux). The project explores symbiotic child-robot interactions in joint cooperative activities, with application to the treatment of children with Autism Spectrum Disorder impairments. The grantee is expected to contribute in implementation and research tasks related to speech processing in environmentally challenging conditions and with limited adaptation data available.

The research will be supervised by Ramón Fernandez Astudillo, researcher at INESC-ID. The grantee will be integrated in the Spoken Language Systems Lab (http://www.l2f.inesc-id.pt). The candidate should have a Ph.D Degree related to the topic. He/she should have experience with Robust automatic speech recognition and/or keyword spotting, deep neural networks and C/C++, bash scripting and Python programming languages.

The following additional experience is not obligatory but will be considered in the evaluation: Computational Paralinguistics, Python-theano, Knowledge of the Portuguese language, and Missing features / Observation uncertainty techniques.

The application deadline is March 27, 2015.

Further information in:
http://www.eracareers.pt/opportunities/index.aspx?task=global&jobId=55039

Back

Top

Descriptif
-----------

- Sujet : Développement d'un outil générique d'annotation des ressources linguistiques

- Niveau : M2 / dernière année d'école d'ingénieur

- Domaine : informatique

- Période : à partir de mars-avril 2015

- Durée : 6 mois

Contexte
--------
Acteur majeur des technologies de la langue, ELDA (Evaluations and Language resources Distribution Agency) est une PME dont les activités s'articulent principalement autour de la distribution et de la production de ressources linguistiques.  Dans le cadre des projets de production de ressources linguistiques dont elle est en charge, ELDA est souvent amenée à des tâches de collecte, d'annotation de contrôle qualité, de packaging, etc.).  Afin de pouvoir gérer la grande variété des différents protocoles et formats d'annotations, ELDA projette d?intégrer en une seule plate-forme un ensemble d'outils qui permettent la gestion homogène et transparente des ressources linguistiques annotées.  

Dans ce contexte, ELDA souhaite consolider sa plate-forme, qui devrait permettre :
- d'importer des ressources annotées déjà existantes ;
- de créer de nouvelles ressources linguistiques, y compris l'annotation de nouveaux corpus ;
- de naviguer dans une ressource linguistique recensée dans la plate-forme, ou à travers plusieurs ressources ;
- de récupérer et d'effectuer des statistiques sur des éléments des corpus et/ou des annotations.


Travail à réaliser
------------------

Au sein de l'équipe de développement informatique d'ELDA, sous la tutelle d'un ingénieur spécialiste du traitement automatique des langues, vous serez amené à participer aux travaux suivants :

- faire un état de l'art sur les outils d'annotation de ressources linguistiques disponibles actuellement ;

- spécifier / réviser l'architecture de l'outil d'annotation à mettre en place chez ELDA ;

- réaliser des développements dans des outils d'annotation, de concert avec les autres membres de l'équipe de développement informatique d'ELDA ;

- développer des composantes logicielles d'importation et d'exportation de données annotées depuis et vers des formats existants, autres que le format utilisé par l'outil d'annotation.
Vos participerez également aux réunions périodiques de l'équipe de développements logiciels d'ELDA.


Profil souhaité
---------------

- BAC + 4/5 / Dernière année d'École d'ingénieur ;
- Bonnes connaissances pratiques de la programmation orientée objet ;
- Connaissances de base en algorithmique ;
- Connaissances de base des architectures des applications Web ;
- Connaissance pratique d'un système de gestion de bases de données ;
- Anglais technique ;
- La connaissance du langage Python sera appréciée ;
- Des notions sur le langage JavaScript seront un plus.


Candidature
-----------

Ce stage, d'une durée de 6 mois et basé à Paris dans le 13e arrondissement (Les Gobelins), est à pourvoir au printemps 2015.

Les candidatures (CV, lettre de motivation) doivent être adressées à Vladimir Popescu (vladimir@elda.org).

Le stage fait l'objet d'une rémunération, variable en fonction du niveau d'études du candidat.

www.elda.org
 

Back

Top

Postdocs in Robust Speaker Verification, Aalborg University, Denmark

At the Faculty of Engineering and Science, Department of Electronic Systems, The section for Signal and Information Processing, two postdoc positions in Robust Speaker Verification are open for appointment from May 1, 2015, or as soon as possible thereafter. The positions are available for a period of 2 years. The Department of Electronic Systems is one of the largest departments at Aalborg University with a total of more than 300 employees. The department is internationally recognized in particular for its contributions within Information and Communication Technology (ICT). The department focuses on maintaining a close interplay with the university's surroundings - locally, nationally and internationally ? as well as producing unique basic research and educating talented and creative engineers. The department collaborates with leading ICT researchers all over the world.

The two postdocs will work on a Horizon 2020 project funded by the European Commission. This project aims at developing innovative speaker verification methods for access control. This relies on developing the core technologies for both text-independent and text-dependent speaker verification and on developing methods robust to varying acoustic environments and threats from spoofing. The results will be used to build a robust speaker verification platform for data-sensitive and mission-critical applications. The applications are from various business sectors and presented by the industrial partners in the project consortium.
The postdocs will cooperate closely with a number of leading academic and industrial partners from the European countries. The successful applicants must have a PhD degree in speaker recognition, noise-robust speech processing, statistical signal processing or machine learning and preferably have experience with developing speaker recognition algorithms or robust speech systems.

You may obtain further information from Associate Professor Zheng-Hua Tan, Department of Electronic Systems, phone: +45 9940 8686, email: zt@es.aau.dk, concerning the scientific aspects of the position.

Application deadline: April 15, 2015.

Webpage for the position: http://www.vacancies.aau.dk/show-vacancy/?vacancy=722471

Back

Top

L’Ircam recrute pour son département Recherche et Développement,

Un(e) chargé(e) de recherche en séparation de signaux de parole appliquée à la robotique

en contrat à durée déterminée à temps plein de 11 mois (prolongation d’un mois possible)

à compter de mai 2015

Présentation de l'IRCAM

L’IRCAM est une association à but non lucratif, associée au Centre national d'art et de culture Georges Pompidou,

dont les missions comprennent des activités de recherche, de création et de pédagogie autour de la musique du

XXème siècle et de ses relations avec les sciences et technologies. Au sein de son département R&D et de son

UMR 9912 STMS commune à l’Ircam, à l’Université Pierre et Marie Curie (UPMC) et au CNRS, des équipes

spécialisées mènent des travaux de recherche et de développement informatique dans les domaines de

l'acoustique, du traitement des signaux sonores, des technologies d’interaction, de l’informatique musicale et de la

musicologie. L'IRCAM est situé au centre de Paris à proximité du Centre Georges Pompidou au 1, Place Stravinsky

75004 Paris.

Présentation de l'ISIR

L'Institut des Systèmes Intelligents et de Robotique (ISIR) est un laboratoire de recherche pluridisciplinaire qui

rassemble des chercheurs et enseignants-chercheurs relevant de différentes disciplines des Sciences de l’Ingénieur

et de l’Information ainsi que des Sciences du Vivant.

L’ISIR est une Unité Mixte de Recherche (UMR7222) commune à l’Université Pierre et Marie Curie (UPMC) et au

Centre National de la Recherche Scientifique (CNRS). L'ISIR est rattaché d’une part à la faculté d’Ingénierie de

l’UPMC (UFR 919) et d’autre part à l’Institut des Sciences de l'Information et de leurs Interactions (INS2I) du CNRS.

L’Institut national de la santé et de la recherche médicale (INSERM) est également tutelle de l'une de ses équipes,

l’Equipe de recherche labellisée (ERL) U1150.

Introduction au projet ROUTE

Le projet ROUTE se positionne dans le cadre d'une collaboration entre l'Institut de Recherche et Coordination

Acoustique Musique (IRCAM) et l’Institut des Systèmes Intelligents et de Robotique (ISIR). Le cadre scientifique du

projet est l’analyse automatique de scène sonore pour l’écoute artificielle d’un robot et l’interaction homme/robot.

L’objectif scientifique du projet est de spécifier un formalisme de décomposition non-négative d’un signal audio pour

la détection et la localisation d’un locuteur dans une scène sonore binaurale. Le système devra permettre à un robot

de détecter la présence et la position d’un locuteur dans un environnement sonore bruité (bruit interne du robot et

bruit externe de la scène sonore). L’objectif applicatif du projet est d’améliorer l’écoute artificielle d’un robot pour

l’interaction homme/robot (identification de locuteur, reconnaissance de parole, etc…).

Missions

Le/la chercheur/se recruté/e sera en charge de l'extension d’un système d’analyse automatique de scène sonore

pour la détection et la localisation de locuteur dans un contexte de captation sonore binaurale.

Il/elle sera en charge de la réalisation d’un système de détection et de localisation robuste au sein d’un

environnement sonore bruité (bruit interne du robot et bruit externe de la scène sonore).

Il/elle participera à l’élaboration d’un scénario d’usage réaliste, et à toutes les activités liées au projet (spécifications,

réunions, expérimentations, évaluations, production des rapports).

Profil recherché

- Excellente connaissance et expérience de décomposition non-négative, particulièrement dans le cas

multicanal. Une thèse dans le domaine est fortement souhaitée.

- Excellente connaissance en traitement du signal (analyse spectrale, extraction de descripteurs audio,

estimation de paramètres)

- Fort intérêt pour la robotique et l'interaction homme/robot.

- Excellente connaissance du langage Matlab, expérience du langage Python (NumPY/SciPY), compétences de

programmation en C/C++.

- Bonne connaissance des environnements Linux, et/ou Mac OS X.

- Productivité élevée, capacité de travailler de façon méthodique et autonome, excellent style de programmation,

créativité, bonnes aptitudes en communication, rigueur.

Salaire

Selon profil et expérience professionnelle.

Candidatures

Prière d'envoyer une lettre de motivation et un CV détaillant le niveau d'expérience/expertise dans les domaines

mentionnés ci-dessus (ainsi que tout autre information pertinente) avant le 15 avril 2015 à roebel_at_ircam_dot_fr

Researcher (M/W) in speech separation applied to robotics (Emergence project: ROUTE)

Availability: early May 2015 Duration: 11 months (possible extension to 12 months)

Introduction to IRCAM

IRCAM is a non-profit organization that is associated to the Centre Pompidou (Centre national d’art et de culture

Georges Pompidou). Its missions comprise research, production, and education related to contemporary music and

its relation to science and technology. Within the R&D department specialized teams are conducting research and

development in the areas of acoustics, sound signal processing, interaction, computer music and musicology.

IRCAM is located in the centre of Paris near the Centre Pompidou, at 1, Place Igor Stravinsky 75004 Paris.

Introduction to ISIR

The Institute for Intelligent Systems and Robotics (ISIR) is a multidisciplinary research laboratory that brings together

researchers and academics from different disciplines of Engineering Sciences and Information and the Life Sciences.

ISIR is a joint research laboratory between the Université Pierre et Marie Curie (UPMC) and the Centre National de

la Recherche Scientifique (CNRS).

Introduction to ROUTE project

ROUTE is a collaborative program directed by the Analysis/Synthesis team at IRCAM and ISIR. The program is

focused on the automatic analysis of an acoustic scene, based on a non-negative decomposition of a multi channel

recording of the scene. This research problem is closely related to several applications in sound processing (e.g.

source separation, CASA); and it is becoming as well an important problem for many applications emerging from the

fields of robotics and artificial audition (e.g. guide robots, older adults monitoring, and the general study of

human/robot interaction).

The principal aim of the project is to evaluate recent advances in non-negative signal decomposition for detecting

and localizing a speaker in a binaural acoustic scene: this study shall allow a robot to identify speech in a

background noise, and localise the speaker's position.

Position description

The hired researcher will be in charge of the extension of the existing analysis system to speaker detection and to

binaural capture. He/she will realize a study on the impact of the noise made by the robot on the performance of the

system. He/she will participate to the elaboration of a simplified realistic test, and to the evaluation of the results, as

well as to all activities related to the project (evaluations, meetings, specifications, reports).

Required Experiences and Skills

- Excellent knowledge and experience with non negative signal decomposition, especially from multi-channel

audio. A PhD in this domain will be highly appreciated.

- Excellent knowledge and experience in audio signal processing (spectral analysis, audio-feature extraction,

parameter estimation)

- High interest in robotics and human/robot interaction

- Very proficient in Matlab, experience with Python (NumPY/SciPY), skills in C/C++ programming.

- Good knowledge of Linux, and/or Mac OS X.

- High productivity, capacity for methodical and autonomous work, creativity, good communication skills, rigor,

and excellent programming style.

Salary

According to professional experience.

Applications

Please send an application letter before 15th April 2015 together with your resume and any suitable information

addressing the above issues preferably by email to: roebel_at_ircam_dot_fr

Back

Top

Post-doctoral Position (24 months)

University of Michigan

Predicting Individual Outcomes for Rapid Interventions

(PRIORI)

Context: Bipolar disorder (BP) is a common and severe psychiatric illness characterized by

pathological swings of mania and depression and is associated with devastating personal,

social, and vocational consequences (suicide occurs in up to 20% of cases [1]). Bipolar disorder

is among the leading causes of disability worldwide [2]. The cost in the United States alone was

estimated at $45 billion annually [3]. These economic and human costs, along with the rapidly

increasing price of health care provide the impetus for a major paradigm shift in health care

service delivery, namely to monitor and prioritize care with a focus on prevention. In this project

we focus on how to using mobile health technologies to automatically track mood variation in

populations with BP. We have a large collection of ecologically recorded smartphone calls, now

numbering over 22,000, and are investigating how to characterize mood in these data.

Overview of Position: The candidate will have the opportunity to analyze spoken data from

individuals with bipolar disorder, collected over the period of six months to a year. Our goal is to

predict mood states from natural speech. Significant opportunities exist in machine learning,

speech processing, user personalization, and health-centered computing. We are currently

processing our massive ecologically valid set of speech data from individuals with bipolar

disorder to uncover the acoustic characteristics of speech that will allow us to intuit mood state

(manic, depressed, euthymic) and mood state change. The candidate will interact with an

established team including experts in affective processing (Emily Mower Provost), reinforcement

learning (Satinder Singh), and psychiatry (Melvin McInnis). We are seeking postdoctoral

candidates with experience in machine learning, speech processing , and human-centered

computing.

Keywords: Machine learning, speech processing, and human-centered computing.

Prerequisite: PhD degree in computer science, signal processing, or machine learning. Strong

knowledge in programming languages and database management are desired.

To apply: Interested applicants should email their CV, a cover letter outlining their experience

and research background, two references, and their two-three most relevant papers to Emily

Mower Provost (emilykmp@umich.edu).

References:

[1] Frederick K Goodwin and Kay Redfield Jamison, Manic-depressive illness: bipolar disorders

and recurrent depression, vol. 1, Oxford University Press, 2007.

[2] Alan D Lopez, Colin D Mathers, Majid Ezzati, Dean T Jamison, and Christopher JL Murray,

“Global and regional burden of disease and risk factors, 2001: systematic analysis of population

health data,” The Lancet, vol. 367, no. 9524, pp. 1747–1757, 2006.

[3] Leah S Kleinman, Ana Lowin, Emuella Flood, Gian Gandhi, Eric Edgell, and Dennis A

Revicki, “Costs of bipolar disorder,” Pharmacoeconomics, vol. 21, no. 9, pp. 601–622, 2003.

Back

Top

Efforts and coordination of speech gestures

Disciplines: Acoustic phonetics, Biomechanics, Physiology, Cognition

Laboratory: GIPSA-lab, Grenoble

Supervision: Maëva Garnier, Pascal Perrier, Franck Quaine

Duration: 3 years (starting between Sept and Dec 2014)

Contact: maeva.garnier@gipsa-lab.grenoble-inp.fr / +33 4 76 57 50 61

Context: This Ph.D thesis takes part to the ANR project StopNCo, dealing with the characterization

and understanding of the physiological efforts and the gesture coordination in speech production1.

Stop consonants (/p/, /t/, /k/, /b/, /d/ or /g/) are of particular interest for the study of speech motor

control, as they require a precise coordination of breathing, laryngeal and articulatory gestures in their

force and timing.

General questions: Stop consonants

are created by an occlusion of the vocal

tract that can occur at 3 different “places

of articulation” in French: at lips (for /p/

and /b/), just behind the teeth (for /t/ and

/d/) or at the back of the palate (for /k/

and /g/) (see Figure). The release of this

occlusion creates a short explosion noise

(or “burst”) and a quick variation in

frequency of the vocal tract resonances

(“formant transients”). These acoustic

features differ significantly between the 3

places of articulation

The objectives of the project are to characterize and to model:

1. by which coordination of breathing and articulatory gestures we control the finer variation of these

acoustic cues (burst spectrum and formant transients)?

2. how these cues are modified when speakers speak more clearly and try to enhance the perceptual

contrast between these 3 places of articulation?

3. how this control develops in children and can dysfunction in some of them?

4. how this control can vary in efficiency, i.e. in the ratio between the acoustic outcomes and the

physiological efforts?

Program:

The first step of the project will consist in implementing new methodologies to measure lip and

tongue articulation efforts, using surface electromyography (EMG), force sensors and electromagnetic

articulography (EMA) (see next figure).

1

see http://www.agence-nationale-recherche.fr/en/anr-funded-project/?tx_lwmsuivibilan_pi2%5BCODE%5D=ANR-14-CE30-

0017

Multiple EMG electrodes will be placed

around the lips to characterize the muscle

activity in different speech movements

and to find global descriptors of the

degree of articulation effort and fatigue.

Force sensors will be used, searching their

optimal number and position on the lips

and palate. We will also try to characterize

the tongue and lips stiffness in order to

take account of it in the calibration of the

force measurements.

Finally, the articulation force estimated by

these two methodologies will be

confronted to the velocity peaks of

tongue and lip movements measured with

EMA, as well as to the perceptual selfevaluation

of the effort spent by the

subject.

The Ph.D thesis will base on these methodologies to characterize the coordination of breathing,

laryngeal and articulatory gestures in the production of stop consonants in healthy adult speakers. A

large database will be recorded with synchronous physiological and acoustical signals, on several

speakers, in controlled laboratory conditions, and for a variety of voice qualities and efforts (whisper to

shout, slow to fast speech rate, etc.). Using statistical data processing and mapping techniques, you

will establish a functional model able to predict the variation of acoustical outcomes from the covariation

of physiological parameters.

In a second step, a second experiment will be conducted in a more realistic and interactive situation

of face-to-face communication. You will explore how speakers modify their production of stop

consonants when they communicate in noisy or reverberant environments, and how they consequently

modify the coordination and the effort of their speech gestures, in comparison to casual speech.

Collaborations

The project will take place at GIPSA-lab in Grenoble, under the co-supervision of Maëva Garnier

(Expert in speech and cognition), Pascal Perrier (Expert in Speech motor control) and Franck Quaine

(Expert in biomechanics and EMG signals), in close relationship with the medical field (a dentist and a

maxillofacial surgeon).

The Ph.D thesis will belong to a larger project, involving a second team that works on laryngeal efforts

(including a ENT specialist), and a third team working on the development of this coordination in

children (including speech therapists from Grenoble’s hospital). These two teams will use the

methodologies developed during the beginning of the Ph.D. thesis, and will bring complementary

information to the functional model of stop consonants production.

During the Ph.D. thesis, we envisage to send the Ph.D candidate for about 3 months in Italy for a

collaboration on high-density EMG matrices.

Skills: We are looking for an open-minded student, coming from either an engineering background

(electronics, mechanics, acoustics) or a human sciences background (phonetics, cognitive sciences),

but having in any case at least some basic technical skills (programming, signal and data processing)

and a strong interest in human-related questions (physiology, speech, human interaction). An

experimental approach will be appreciated.

Indemnities: 1400 € per month during the 3 years of Ph.D fellowship. The Ph.D candidate will be

able to have additional teaching activies.

Back

Top

Role: Junior Computational Linguist

About RecordSure

RecordSure builds leading global, first tomarket

products for the compliance auditing of Financial Services

conversations.

RecordSure was formed in 2012 and is a ‘sister company’ of The Consulting Consortium, the UK’s leading

independent consultancy providing compliance services to the financial services sector, both nationally and

globally.

Primarily privately funded, both The Consulting Consortium and RecordSure are also inpart

owned by the

Business Growth Fund, backed by Barclays, HSBC, Lloyds, RBS and Standard Chartered banks.

Our Culture

We are a dynamic start up, who, although still relatively small, have gained significant traction at board level

in many of the world’s largest financial service institutions.

Passionate about technology, we love solving hard problems; and as individuals we are each as eager to

learn as we are to help each other to do so.

RecordSure is committed to safeguarding our Client Data and our Intellectual Property and expects all staff

to share this commitment.

The Role

RecordSure requires a Junior Computational Linguist to join our Language Science Team working on cuttingedge textanalytics

and machine learning solutions in order to enhance the RecordSure platform.

The role includes a very close collaboration with the Speech and Data science teams, as well as with the

development team and would suit a recent graduate.

The role revolves around building, refining and maintaining a knowledge base of speech within the finance

industry and requires a candidate who has experience working with lexical databases.

Experience in Artificial Intelligence, Machine Learning, and Information Extraction is desired, but not

required. Basic programming skills are essential.

Responsibilities:

● Building and maintaining a lexical database

● Working alongside the development team to build tools for database management

● Extracting, adapting and evaluating domain specific terminologies

● Be aware of the company processes relating to ISO 27001 and maintain an awareness of the

importance of secure handling of Client data and RecordSure Intellectual Property

● Comply with internal policies, in particular raising any concerns over security issues

Your Attributes

You’re an intelligent, enthusiastic and highly motivated handson

candidate with a solid base in coding.

You are driven by:

● Learning about new technologies every day

● Solving challenging problems

● Quality people in your team

The Candidate

● BSc/MSc degree or equivalent in Computational Linguistics, Computer Science or related discipline

● Experience in corpus building

● Experience in phonetic transcriptions and localization process

● Experience using scripting languages for manipulation of data e.g. Shell, Python

Contact: ibrahim.almajai@recordsure.com

Back

Top

Une offre de thèse de doctorat est disponible dans l'équipe Synalp à Nancy sur l'analyse sémantique de Twitter:

• Duration: 3 years, starting from October 2015
• Salary: 1760 euros, charges included
• Application deadline: 4 May 2015

Plus d'informations: http://talc1.loria.fr/synalp/posts/phd-twitter.html

Contact: cerisara@loria.fr

Back

Top

Une offre de thèse de doctorat (3 ans, à partir d?octobre 2015) est disponible dans l'équipe EXPRESSION de l?IRISA à Lannion sur la détection multimodale d?anormalité dans le discours : utilisation de la voix et des expressions faciales.

Plus de détails : le sujet (en Anglais)

Contact : arnaud.delhay@irisa.fr, pierre-francois.marteau@irisa.fr, damien.lolive@irisa.fr

Back

Top

Assistant Professor in Socially Aware ICT

KTH – Royal Institute of Technology invites applications for a tenure-track faculty appointment in Socially Aware ICT at the rank of Assistant Professor. We seek individuals with outstanding potential for research achievement, and a strong interest and commitment to undergraduate and graduate teaching. We seek candidates that can add to or extend the current research areas at KTH with more emphasis being put on excellence than on candidates’ specific area of expertise. The position comes with a significant start package that will allow the successful candidate to quickly establish their own research group.

KTH carries out a wide range of internationally renowned research related to Socially Aware ICT.

Research groups working in social robotics, computer vision, speech technology and human–computer interaction, comprise part of the School of Computer Science and Communication. For more information about the school, see www.kth.se/csc

Research groups in technology in caring sciences and health systems engineering comprise part of the School of Technology and Health. For more information about the school, see www.kth.se/sth

For further information and details on application procedure see www.kth.se/applytoassistantprofessor

Application deadline is May 29, 2015.

Applicants having earned their PhD no later than May 2008 will be given priority. The term of appointment is expected to begin around November 2015, or later depending on availability.

Contacts:

Prof Danica Kragic, Head of Centre for Autonomous systems, danik@csc.kth.se

Prof Joakim Gustafson, Head of department of Speech, Music and Hearing, jocke@speech.kth.se

Prof Britt Östlund, School of Technology and Health, brittost@kth.se

Ass Prof Hedvig Kjellström, Head of the Computer Vision and Active Perception Lab, hedvig@csc.kth.se

Prof Kia Höök, Head of MobileLife, khook@kth.se

Prof Roberto Bresin, Department of Media Technology and Interaction Design, oberto@kth.se

Back

Top

Postdoctoral research position:

Wordspotters for radiomining in rural Africa

A postdoc position focussing on automatic identification of spoken keywords in multilingual environments with few or no resources is available in the Digital Signal Processing Group of the Department of Electrical and Electronic Engineering at the University of Stellenbosch. The project will develop wordspotters that can be used to monitor community radio broadcasts in rural African regions as a source of early warning information during natural disasters, disease outbreaks, or other crises. Specific project objectives include the development of a research system, the production of associated publishable outputs, and the development of a web-based demonstrator. The position is part of a collaborative project with the United Nations Global Pulse. Further information is available on the web at http://pulselabkampala.ug/radiomining.

Applicants should hold a PhD (preferably obtained within the last 5 years) in the field of Electronic/Electrical Engineering, Information Engineering, or Computer Science, or other relevant disciplines. Suitable candidates must have practical experience with automatic speech processing systems and should have an excellent background in statistical modelling, signal processing, and/or speech analysis. The candidate must also have good programming skills and be able to use high level programming languages for developing prototype systems. Finally, candidates must have excellent English writing skills and have an explicit interest in scientific research and publication.

The position will be available for one year, with a possible extension to a second year, depending on progress and available funds.

Applications should include a covering letter, curriculum vitae, list of publications, research projects, conference participation and details of three contactable referees and should be sent as soon as possible to: Prof Thomas Niesler, Department of Electrical and Electronic Engineering, University of Stellenbosch, Private Bag X1, Matieland 7602. Applications can also be sent by email to: trn@sun.ac.za. The successful applicant will be subject to University policies and procedures.

Interested applicants are welcome to contact me at the above e-mail address for further information regarding the project.

Back

Top

Postdoc position at Aix-Marseille University (France) in opinion
analysis and novelty discovery

Application deadline: 04/31/2015
Starting: as soon as possible.

Description:

The ADNVIDEO project, funded in the framework of A*MIDEX
(http://amidex.univ-amu.fr/en/home), aims at extending multimodal
analysis models. It focuses on jointly processing audio, speech
transcripts, images, scenes, text overlays and user feedback.

Using as starting point the corpus, annotations and approaches
developed during the REPERE challenge (http://defi-repere.fr), this
project aims at going beyond indexing at single modalities by
incorporating information retrieval methods, not only from broadcast
television shows, but more generally on video documents requiring
multimodal scene analysis. The novelty here is to combine and
correlate information from different sources to enhance the
description of the content. In particular, the use case considered may
have, in terms of technology transfer, significant economic benefits
regarding automatic information monitoring of people or brands:
automatically find the most relevant content about a person or a brand
and if this content is positive or not.

Objectives:

The candidate will participate in the development of a prototype for
video monitoring, leading to a technology transfer towards business:
* Extraction of multimodal high-level descriptors (semantic-oriented
descriptors). These semantic-oriented descriptors are extracted from
low-level descriptors, especially from text transcripts.
o Extract concepts (conceptualization) from the transcript of the audio.
o Detection of semantic chapters by aggregation of concepts extracted
shot/scene-wise
o Detection of novelty about peoples or brands
o Detection and extraction of sentiments and opinions
* Aggregation of multimodal descriptors to form the multimodal
footprint of a video.
* Matching videos and promotional material using recommendation approaches
* Validation of the video recommendation prototype.
* Participation to the scientific life of the lab, including paper publication.
The allocation of the tasks can be adjusted depending on the wishes
and skills of the candidate.

Skills:

For this project, we are looking for one candidate with a PhD degree
in the areas of information retrieval, natural language processing,
machine learning:
* Strong programming skills (C++, Java, Python?).
* Desire to produce functioning end-to-end systems, life-scale live demos
* Scientific rigor
* Imagination
* Top notch publications
* Excellent communication skills
* Enjoy teamwork

Location:

The work will be conducted in the University of Aix-Marseille at the
Laboratoire des Science de l?information et des système (LSIS
http://www.lsis.org), within the ADNVidéo project, supported by
funding from a AMIDEX foundation in collaboration with Kalyzee
(http://www.kalyzee.com/). Both LSIS and Kalysee are located in the
historical and sunny city of Marseille, in south of France
(http://www.marseille.fr/sitevdm/versions-etrangeres/english--discover-marseille).

Contact: sebastien.fournier@lsis.org
Duration: 23 month
Candidates should email a letter of application, a detailed CV
including a complete list of publications, and source code showcasing
programming skills. ---

 

Back

Top

Sujet : Espaces de représentation sémantique distribues pour les tâches de traduction automatique (compréhension et génération de la parole) dans les systèmes d'interaction vocale

Keywords: word embeddings, deep neural networks, statistical machine translation, spoken language understanding, natural language generation

Description : L’émergence de l’accès universel à la société numérique implique le développement de systèmes d'interaction multilingues : par exemple pour la recherche d'information dans des corpus multimédia multilingues ou pour le développement de systèmes de dialogue multilingues (dont les tâches peuvent aller des systèmes de question/réponse jusqu'à la résolution de problèmes complexes, comme l'aide aux usagers d'une compagnie). Dans ce contexte la traduction automatique n'est pas limitée au passage entre deux langues humaines, ainsi la compréhension et la génération de la parole peuvent être vues comme des exemples de tâches de traduction de la parole et seront étudiées dans le cadre de cette thèse. La recherche d'une solution optimale pour l'ensemble des tâches visées sera bien sur un élément clef de cette étude.

Depuis une dizaine d’années, les approches les plus performantes pour la traduction automatique sont basées sur l’utilisation de modèles probabilistes. Pour être efficaces, de telles approches nécessitent de disposer de larges bases de données d'exemples (dans ce cas, des corpus de phrases parallèles entre les langues source et cible), ce qui n'est pas toujours possible, en particulier dans les domaines spécialisés. par ailleurs, dans le cas de la traduction de la parole, les systèmes doivent baser leurs hypothèses sur les sorties imparfaites des systèmes de reconnaissance de parole. Il est donc important de baser la décision sur un maximum d'informations (et pas uniquement sur l’identité des mots présents).

Des approches récentes ont montre l’intérêt d’intégrer l'information sémantique pour réaliser la traduction automatique de la parole par des méthodes statistiques [1]. Les gains en performance restent toutefois limités et une grande marge d’amélioration est encore possible. De plus, la nécessité d'analyser au préalable le texte réduit les possibilités d'application de ces approches dans le contexte des systèmes d'interaction vocale, où l’étape de reconnaissance de la parole en diminue la faisabilité. En effet, l'extraction fine de caractéristiques est fortement perturbée par le niveau élevé de bruit dans les textes à traiter issus d'une étape de décodage automatique.

Aussi l’émergence de nouvelles approches fortement automatiques pour la représentation des données textuelles, par exemple à l'aide de réseaux neuronaux profonds [2], présente une nouvelle opportunité pour développer des approches permettant d'envisager l'utilisation de nouveaux paramètres sur une grande échelle pour guider et améliorer la traduction par la prise en compte renforcée d'information syntactico-sémantiques [3]. Dans le contexte des systèmes de dialogue homme-machine il sera possible d'évaluer la pertinence des approches envisagées sur des tâches et des corpus de tailles contrôlables.

Références :
[1] Dekai Wu et Pascale Fung, Can Semantic Role Labeling Improve SMT?, EAMT, 2009
[2] T. Mikolov, I. Sutskever, K. Chen, G. S Corrado, and J. Dean, “Distributed representations of words and phrases and their compositionality,” in Advances in Neural Information Processing Systems, 2013.
[3] David Chiang, Kevin Knight et Wei Wang, 11,001 New Features for Statistical Machine Translation, NAACL-HLT, 2009

Pré-requis : Master en informatique avec une composante sur les méthodes d'apprentissage automatique et/ou sur l’ingénierie de la langue

Encadrant : Prof. Fabrice Lefèvre (co-encadrants : Stéphane Huet et Bassam Jabaian)
Lieu : LIA-CERI-Univ. Avignon
Candidature : envoyer un mail avec un CV, une lettre de motivation et vos relevés de notes avant le 12 mai à fabrice.lefevre@univ-avignon.fr en mettant en copie  bassam.jabaian@univ-avignon.fret Stephane.Huet@univ-avignon.fr

Back

Top

TWO POST-DOCTORAL POSITIONS FOR THE PROJECT IFCASL (LORIA, NANCY, FRANCE)

We invite applications for a one-year Post?Doctoral position at the Laboratoire LORIA (Lorraine university, INRIA, CNRS, UMR 7306, France), to work on the project IFCASL (Individualised Feedback in computer-Assisted Spoken Language learning).

?    Description
The IFCASL project is funded by the Deutsche Forschunggemeinschaft (DFG) and the French National Research Agency (ANR). It aims at increasing the usability and efficiency of computer-assisted language learning software by adapting content, feedback and exercises to individual learners in the speech dimension of a foreign language. This will be achieved by integrating the production and perception of the learner?s own speech and focusing on the French-German language pair, in both directions.

The objective is to offer feedback that relies on phonetic knowledge incorporated in the learning system and interacts relevantly with automatic speech recognition (ASR) and signal processing. This means that confidence in the results returned by ASR will be taken into account to derive diagnoses and feedback: the higher the confidence the more explicit the level of feedback.

A corpus of German recorded by French learners and a second of French recorded by German learners will be exploited to determine errors on the one hand, and to develop algorithms involved in feedback on the other hand.

?    jobdescription
The post?doctoral fellow will be mainly involved in the development of the signal processing and automatic speech recognition algorithms intended to realize the diagnosis and acoustic feedback provided to learners. 

?    Qualifications
A Ph.D. in signal processing or in computer sciences (automatic speech recognition, statistic modeling) and solid competence/experience in JAVA development are required. Experience in phonetics and speaking German and French are also welcome.

?    Application procedure
Candidates should send a detailed CV with a list of publications, and a cover letter with statement of research interests and details of their experience in signal processing, automatic speech recognition and JAVA.
Please e-mail documents to: Yves.Laprie@loria.fr (Yves Laprie, LORIA, Nancy, France).

Deadline for submission: June 15, 2015
Expected start date: September 2015 (There is some flexibility.)
Length of contract: 12 months
Salary: ?2000/month (net)

Back

Top

PhD Position

IRISA lab, University of Rennes 1 at Lannion, Côtes d’Armor

Expression Team http://www-expression.irisa.fr/

Subject: multimodal detection of abnormality in discourse: using voice and facial expressions

Application: URGENT

Please send a CV and reference letters by e-mail to all the following contacts: Arnaud Delhay

(arnaud.delhay@irisa.fr), Pierre-François Marteau (pierre-francois.marteau@irisa.fr) and Damien

Lolive (damien.lolive@irisa.fr) BEFORE the 4th of May 2015.

The thesis will be co-funded by the DGA (French Defence ministry). The candidate must have the

nationality of a country of the European Union or of Switzerland. S/he must hold a Master

degree (or equivalent) in computer science.

The candidate is expected to conduct cutting-edge applied research in one or several of the

following domains: signal processing, statistical machine learning, speech and gesture recognition.

S/he should have excellent computer programming skills (e.g. C/C++, Python/Perl, etc.), and

possibly knowledge in machine learning, signal processing or human computer interaction.

Duration: 3 years

Date: October 2015 – September 2018

This PhD, proposed by the EXPRESSION team at IRISA, will address the detection of abnormality

from facial movements and speech signals of a human being in a situation of stress. We mean by

‘abnormality’ existence of foreign elements to a normal situation in a given context. The study will

focus in particular on the joint use of facial and vocal expression parameters to detect abnormal

variations of expressivity in speech, not only related to emotion, but also to social interactions and

psychological signals. These abnormal signals can appear in extreme stress situations for pilots or

vehicle drivers, for example. This study could also find applications in the medical field, e.g.,

detection of abnormal behaviors due to mental disabilities such as autism.

We aimed at developing a system capable of detecting abnormal behaviors by the analysis of

records of concrete situations. The thesis will then explore several issues including the followings:

• Collect, segment and annotate multimodal data;

• Identification of descriptors enabling the description of abnormality;

• Development of dedicated machine learning approaches for abnormality detection;

• Development of a decision system.

Keywords: Speech, facial expressivity, gesture analysis, heterogeneous information, machine

learning, classification

IRISA/MID/EXPRESSION 2/2

Bibliography:

[1] Carlos Busso, Zhigang Deng, Serdar Yildirim, Murtaza Bulut, Chul Min Lee, Abe Kazemzadeh,

Sungbok Lee, Ulrich Neumann, and Shrikanth Narayanan. Analysis of emotion recognition using

facial expressions, speech and multimodal information. In Proceedings of the 6th international

conference on Multimodal interfaces, pages 205–211. ACM, 2004.

[2] B. Fasel and Juergen Luettin. Automatic facial expression analysis : a survey. Pattern Recognition,

36(1) :259 – 275, 2003.

[3] Wesley Mattheyses and Werner Verhelst. Audiovisual speech synthesis : An overview of the

state-of-the-art. Speech Communication, 66(0) :182 – 217, 2015.

[4] Marie Tahon. Acoustic analysis of speakers emotional voices during a human-robot interaction.

Theses, Université Paris Sud - Paris XI, November 2012.

[5] Mariette Soury. Multimodal stress detection for remediation software design. Theses, Université

Paris Sud - Paris XI, October 2014.

[6] Soujanya Poria, Erik Cambria, Amir Hussain, and Guang-Bin Huang. Towards an intelligent

framework for multimodal affective data analysis. Neural Networks, 63(0) :104 – 116, 2015.

[7] D Govind and SR Mahadeva Prasanna. Expressive speech synthesis : a review. International

Journal of Speech Technology, pages 1–24, 2013.

[8] Marc Le Tallec, Jeanne Villaneau, Jean-Yves Antoine, Agata Savary, and Arielle Syssau-

Vaccarella. Emologus - a compositional model of emotion detection based on the propositionnal

content of spoken utterances. In Text, Speech and Dialogue, Proc., Brno, Czech Republic, 2010.

Back

Top

Post-doc position at LORIA (Nancy, France)

Automatic speech recognition: Deep Neural Network for language model

Framework of ANR project ContNomina

The technologies involved in information retrieval in large audio/video databases are often based on the analysis of large, but closed, corpora, and on machine learning techniques and statistical modeling of the written and spoken language. The effectiveness of these approaches is now widely acknowledged, but they nevertheless have major flaws, particularly for what concern proper names, that are crucial for the interpretation of the content.

In the context of diachronic data (data which change over time) new proper names appear constantly requiring dynamic updates of the lexicons and language models used by the speech recognition system.

As a result, the ANR project ContNomina (2013-2017) focuses on the problem of proper names in automatic audio processing systems by exploiting in the most efficient way the context of the processed documents. To do this, the post-doc student will address the contextualization of the recognition module through the dynamic adjustment of the language model in order to make it more accurate.

Post-doc subject

Deep Neural Network have become a key component of modern automatic speech recognition systems. The language model of our recognition system is based on a neural network learned from a large corpus of text. The problem is to estimate the probability of a new proper name depending on its context. Several tracks will be explored: adapting the language model, using a class model or studying the notion of analogy.

Our team has developed a fully automatic system for speech recognition to transcribe a radio broadcast from the corresponding audio file. The post-doc will develop a new module whose function is to integrate new proper names in the language model.

Required skills

A PhD in signal processing or in computer sciences, be familiar with the tools for automatic speech recognition, background in statistics and computer program skills (C, object-oriented programming and Perl).

Post-doc duration: 12 months, start during the 2015 (these is some flexibility)

Localization and contacts: Loria laboratory, Speech team, Nancy, France

irina.illina@loria.frdominique.fohr@loria.fr georges.linares@univ-avignon.fr

Candidates should email a letter of application, a detailed CV with a list of publications and diploma

Back

Top

Post-doctoral Position in Speech Signal Processing

The School of Science and Engineering at Reykjavik University invites applications for a 1 to 2

year post-doctoral position in the area of speech signal processing. The position is available

from summer 2015 or as soon as possible thereafter. The main responsibility of the research will

be to lead the scientific work in the Icelandic Research Fund (IRF) Project “Model-Based Speech

Analysis and Voice Quality Assessment.” Other responsibilities will include mentoring graduate

students and participating in the further development of speech signal processing within

the university.

About the project:

The project aims to combine non-linear signal analysis methods with model-based signal

processing. The immediate aim is to produce a rich description of the speech signal with

respect to voice quality, event markings and signal states. This creates a foundation for a

targeted modeling of speech so that linguistic content, speaker identity, voice quality and affect

can be determined. The applicant is expected to implement and develop state-of-the-art signal

processing and machine learning algorithms, using for example Matlab, Python and Kaldi.

Research environment:

The project is carried out at the Center for Design and Analysis of Intelligent Agents (CADIA) and

it is expected that 1-3 graduate students will be working on parts of the project as

well. The center offers a broad range of skills and knowledge in speech- and language-related

areas spanning cognitive psychology, virtual environments, general artificial intelligence and

automatic speech recognition. Examples of related research projects within the center

include IRF-Projects “Cognitive Workload Monitoring using Speech” and “Virtual Reykjavik” and

two industry projects on automatic speech recognition.

We expect:

• a PhD degree in signal processing, machine learning or related fields. Candidates about to earn

their doctorate are welcome to apply.

• advanced methodological skills: handling data, processing signals and building prototype

software

• high motivation and interest in joining a cutting-edge speech signal processing group

• creativity and interest shaping your own research project

• high-level interpersonal skills, including strong commitment to teamwork

• very good communication skills in English

We offer:

• top-level research facilities in a dynamic and collegiate intellectual environment

• a comprehensive mentoring programme with supervision by world-leading academics

• a salary package range from ISK 4.400.000 – 5.100.000 per year gross salary plus relocation

expences. There is also a travel fund available for conference activities.

• A position based in Reykjavik, located between North America and Europe with excellent flight

connections, a growing innovation and technology sector, as well as vibrant culture scene.

How to apply:

Electronic applications, including a cover letter, a CV and the details of two referees, should be

sent to dr. Jon Gudnason (email: jg@ru.is)

Back

Top

PhD thesis: Audio-visual intelligent sensing

Queen Mary, university of London, UK/ Fondazione Bruno Kessler, Trento, Italy

To start around September 2015

Applications are invited for 4 (four) PhD Studentships to undertake research in the areas of computer vision and audio processing for people monitoring in multi-camera multi-microphone environments, and will be part of an interdisciplinary project on audio-visual monitoring for smart interactive and reactive environments. The Studentships (to be started in or after September 2015) are part of an interdisciplinary project between the Centre for Intelligent Sensing (http://cis.eecs.qmul.ac.uk) at Queen Mary University of London (QMUL) and the Centre for Information Technology (http://ict.fbk.eu) at the Fondazione Bruno Kessler (FBK), Trento, Italy.

The Project will focus on methods for long-term people tracking, activity recognition, acoustic scene analysis, behaviour analysis, distant-speech recognition and understanding applied to individuals as well as groups. Such information will enable learning 'patterns of usage' of the environment, and patterns can in turn be used to adapt and optimise the sensing accordingly.

Each PhD student will spend approximatively 50% of their time in London and 50% of their PhD time in Trento and will have access to state-of-the-art audio-visual laboratories, including a multi-camera multi-microphone installation at a large open hallway and a smart home facility equipped with cameras, microphones and automated devices.

More information on the call can be found at http://www.jobs.ac.uk/job/AKZ561/phd-studentships-in-audio-visual-intelligent-sensing

For what concerns activities related to audio and speech processing, please also refer to the following web sites http://shine.fbk.eu and http://dirha.fbk.eu 

Back

Top

Toshiba Research Europe Ltd, UK Laboratory

Speech Technology Researcher

jobDescription

Toshiba Research Europe Ltd (STG-CRL), in Cambridge UK, is looking for researchers to join its Speech Technology Group for a) Acoustic modelling for speech recognition (primarily) and synthesis b) statistical spoken dialogue.

Outstanding PhD-level candidates at a good level of experience are encouraged to apply. Candidates should be highly motivated, team-oriented and should have the ability to work independently. Strong mathematical background and excellent knowledge in signal processing and speech analysis algorithms are required. Good software engineering skills will be a plus.

About us

The lab in Cambridge, in collaboration with other Toshiba groups and speech laboratories in China and Japan, covers all aspects of speech technology and at many levels: from basic and fundamental research to industrial development. We support our researchers in building their career by encourage them to publish their results and by investing in innovation and creation for addressing real problems in speech and language technology. STG-CRL has also strong connections with EU Universities , especially with the Cambridge University Engineering Department.

Salary

We offer competitive salary and benefit packages with private pension plan, bonuses, private health insurance, BUPA Medical, Life Assurance, Group Income Protection

Opening Date

05/05/2015

Closing Date

05/06/2015

How to apply

To apply send your CV and a cover letter to: stg-jobs at crl dot toshiba dot co dot uk, subject: “Application to Researcher Position”

Back

Top

The Sigmedia Group (www.sigmedia.tv) at Trinity College Dublin, Ireland, has a PhD position in speech processing available for Sept 2015. The topic is paralinguistics, specifically applying signal processing to the measurement of speaker engagement. Full details are here: http://adaptcentre.ie/careers/PhD-capturing-speaker-engagement-with-audio-visual-signal-processing.html

Back

Top

Speech Communication with specialization in phonetically-based interaction analysis

KTH Royal Institute of Technology, KTH computer Science and Communication

KTH Royal Institute of Technology in Stockholm is the largest and oldest technical university in Sweden. No less than one-third of Sweden’s technical research and engineering education capacity at university level is provided by KTH. Education and research spans from natural sciences to all branches of engineering and includes Architecture, Industrial Management and Urban Planning. There are a total of 12,400 full year students at first and second l evels, almost 1,900 active (at least 50 per cent) research students and 5,100 employees. For information about KTH Computer Science and Communication, go to https://www.kth.se/en/csc.

Your work will take place at the department of Speech, Music and Hearing. The department is an internationally recognized research lab in speech technology and modelling of spoken communication. Speech research is focused on human-like communication with machines and robots, based on multimodal information. The research area is truly multi-disciplinary bridging computer science, machine learning, linguistics, and perception and cognition disciplines. The research is based on data-driven methods: collection and refinement of corpora, as well as development of tools and language resources is thus a central part of the work. For more information go to, https://www.speech.kth.se.

Subject field

Speech Communication with specialization in phonetically-based interaction analysis

Subject description

Speech Communication concerns the study of human communication involving speech, in which other signals such as facial and bodily movements also are of great importance. The field includes theories, models and technical systems concerning all parts of the communication chain. Central research topics in the specialization towards phonetic interaction analysis include experimental design and execution of data collections; phonetic, prosodic and multimodal analysis; and modelling of the interactional aspects of spoken dialogue.

Duties

The duties involve research and teaching as a part of the Speech Group at Dept. of Speech, Music and Hearing (TMH) at the School of Computer Science and Communication (CSC). The teaching includes advanced courses, master thesis and PhD supervision. The teaching will not exceed 25 %. The assistant professor will be given opportunity to develop their independence as researcher and gain accreditation that may allow them to take other teaching positions with higher eligibility requirements (see Chapter 4, Section 12 a of the Higher Education Ordinance). Following application, the assistant professor shall be assessed for promotion to associate professor.

Qualifications

Eligibility

A person is eligible for appointment if they have attained a Degree of Doctor or have the equivalent academic expertise. Primarily, persons should have attained a Degree of Doctor or have acquired the equivalent expertise no more than seven years before the end of the application period.

Grounds of assessment

Special importance

Scientific and technical skills within the subject area, documented by publications in international journals and main conferences in the area are of special importance. Documented ability to advance the field through independent scientific research on a high international level is also of special importance.

Significant importance

Ability to establish and develop international cooperations within research and education is of significant importance, as is the applicant’s documented experience of teaching in the subject field and related areas at advanced level. The applicant's long-term potential to develop and establish an independent research activity och continuously renew the research area is also of significant importance.

Of importance

It is of importance that a person who is appointed as assistant professor at KTH has experience from research environments other than KTH, equivalent to a post-doctorate period or doctorate degree from another university. Documented experience from initiating and executing externally financed research projects are also of importance.

Also taken into account

Ability to collaborate with the surrounding society and to disseminate information regarding research and development work will also be taken into account. In addition, the applicant's expertise in developing and leading activities and personnel is taken into acocunt. This includes having knowledge about matters of diversity and equal treatment, with particular focus on gender equality.

Special grounds of assessment for promotion to associate professor

When assessing applications for promotion to associate professor, the applicant's ability to independently initiate and carry out research of high academic quality, published in international publications, and the applicant's ability to obtain financing for research operations will be assessed. One special ground for assessment is the applicant's ability to independently establish new collaborations and research specializations. Special consideration will be shown to displayed teaching and supervising expertise.

tradeunion representatives

You'll find contact information to trade union representatives at KTH:s web page.

Application

Your application should follow KTH´s CV template for employment of teachers. You are the main responsible to ensure that the application is complete according to the ad and CV template. Your complete application must be received at KTH no later than the last day of application. Log into KTH's recruitment system in order to apply to this position.

Others

Time limitation: The appointment is for an indefinite term, but no more than four years, and may be extended if, due to the teacher’s absence due to sick leave, parental leave or other special grounds, more time is required to reach the objectives of the appointment. However, the total appointment period may not exceed six years. The appointment is part of the Tenure Track system at KTH and the assistant professor may apply for promotion to tenured associate professor.

We firmly decline all contact with staffing and recruitment agencies and job ad salespersons.

Type of employment: Temporary position longer than 6 months

Working hours: Full time

First day of employment: According to agreement

Salary: Monthly pay. According to agreement.

Number of positions: 1

Working hours: 100%

City: Stockholm

County: Stockholms län

Country: Sweden

Reference number: D-2015-0149

Contact:

1. Joakim Gustafson, Professor TMH, +46 (0) 8 790 8965, jkgu@csc.kth.se

2. Maria Widlund, HR-manager CSC, +46 (0) 8 790 9754, mwidlund@kth.se

3. Kerstin Lagerstedt, queries application procedures, +46 (0) 8-790 7879, human-resources@csc.kth.se

Last application date: 15.Jun.2015

Back

Top

3 Ph.D. and 2 Post-doc positions

at Autonomous University of Zacatecas, Mexico. On the topic of Automatic Speech Recognition and Speech Synthesis for autochthonous Mexican languages. mailto:hamurabigr@uaz.edu.mx (http://piitec.uaz.edu.mx/)

Back

Top

L'équipe LINKMEDIA de l'IRISA recherche un ingénieur R&D spécialiste des technologies
multimédias, Bac+5 ou Bac+8. Rémunération selon niveau d'étude et expérience. Merci de
diffuser auprès des étudiants dans les filières concernées.

=====================================================================
Lieu d?exercice : IRISA, Rennes
Contrat : CDD d?une dure?e de 18 mois de?butant le 1er septembre 2015 Re?mune?ration : de
24 k? a? 35 k? annuels bruts selon expe?rience
Contact : Guillaume Gravier, guillaume.gravier@irisa.fr

L?e?quipe LINKMEDIA de l?IRISA, laboratoire de recherche public, travaille au
de?veloppement de technologies permettant la description et l?acce?s au contenus
multime?dias par analyse de ces derniers : vision par ordinateur, traitement de la parole
et du langage, traitement des contenus audio, fouille de donne?es. Nos recherches portent
par exemple sur des algorithmes de de?tection de noms propres et noms de lieux dans des
textes, de segmentation en locuteur de documents audios, de de?tection d?images
similaires ou encore d?objets particuliers dans les images. Nous e?tudions e?galement
comment assembler ces fonctionnalite?s de base pour organiser des collections de contenus
multime?dias et de?velopper des prototypes d?applications permettant d?exploiter des
collections multime?dias pour en extraire des informations.

Les travaux de l?e?quipe s?appuient sur une plateforme qui a pour objectif d?accompagner
les chercheurs travaillant dans le domaine de l?analyse et de l?exploitation de contenus
multime?dias. La plateforme fournit d?une part une infrastructure mate?rielle pour
acque?rir des donne?es, mettre a? disposition des corpus de re?fe?rence, stocker des
re?sultats de traitements et les partager. D?autre part, nous de?veloppons une offre
logiciel sous la forme de web services en s?appuyant sur une architecture ge?ne?rique et
transparente de mise en ligne de programmes issus de la recherche. Ces services
permettent en particulier un acce?s libre aux technologies multime?dias pour certains
usages : travaux de recherche, test d?ade?quation pour les industriels, travaux pratiques
dans un cadre e?ducatif, etc.

Pour de?velopper et promouvoir les services propose?s sur la plateforme d?indexation
multime?dia de l?IRISA, nous souhaitons recruter un inge?nieur spe?cialiste du traitement
des donne?es multime?dias. Les missions qui lui seront confie?s sont :
? inte?gration a? la plateforme de modules existants
? de?veloppement de nouveaux modules mettant en ?uvre des techniques a? l?e?tat de l?art
? mise en cohe?rence de l?ensemble des modules et documentation
? re?alisation de de?monstrations d?applications multime?dias pour l?e?ducation et le
transfert industriel
? participation a? des campagnes d?e?valuation internationale
L?inge?nieur sera inte?gre? dans l?e?quipe de recherche LINKMEDIA et travaillera en
e?troite collaboration avec les chercheurs et leurs partenaires sur des projets de R&D.

Le candidat, de niveau Bac+5 ou Bac+8, devra posse?der un inte?re?t marque? pour les
technologies multime?dias et les technologies du web. Il devra e?galement justifier d?une
expe?rience significative en programmation (langages C/C++, perl, python), par exemple au
travers de projets et de stages pour les jeunes diplo?me?s. Une expe?rience dans la
conduite de projets informatiques d?envergure sera appre?cie?e. E?tant donne? le contexte
international de travail, une bonne connaissance de l?anglais est indispensable.

Pour candidater, merci d?adresser un CV accompagne? d?une lettre de motivation. Pour plus
de pre?cisions sur le poste, nous contacter.

Back

Top

Trois postes d'ATER en informatique sont disponibles à l'UFR de Sociologie et d'Informatique pour les Sciences Humaines de l'Université Paris Sorbonne avec le profil 'traitement de la parole et du langage'. Les candidats enseigneront l?Informatique dans les différentes formations de licence et de master du département d?Informatique, Mathématiques et de Linguistique appliquées. Ils devront s'inscrire dans un ou plusieurs axes de l'équipe de linguistique computationnelle du laboratoire Sens Texte Informatique et Histoire (STIH)(www.stih.paris-sorbonne.fr/)

La date limite de candidature est le 22 mai 2015 16h
Personne à contacter : Claude.Montacie@Paris-Sorbonne.fr

Back

Top

PhD Research Fellow

Speech Production and Swallowing with Full Dental Prosthesis

(May 28th, 2015)

Context

A position as PhD Research fellow is available in the PCMD (« Perception, Contrôle, Multimodalité et

Dynamiques de la parole ») research team in that GIPSA-Lab. This PhD is part of a research project

involving researchers from two different labs, Gipsa-lab in Grenoble and the LMI in Lyon. It is funded

by the Région Rhône-Alpes for 3 years from next fall, in the context of the research program ARC2

focusing on Quality of Life in the Elderly.

Scientific goals

The general objective of the thesis is to thoroughly study in patients the mechanical interactions

between a full dental prosthesis and the tongue during speech production and swallowing, and to

understand how it influences the control of these basic orofacial motor tasks and its evolution over

time. The work will be organized around three main tasks:

(1) The characterization of the mechanical stress exerted by the tongue on the dental

prosthesis, with a special focus on its amplitude, its spatial distribution and its variation

across patients.

(2) The study of the variation of the mechanical stress during the first weeks following the

insertion of the dental prosthesis in the patient’s mouth, during which a motor adaptation

takes place.

(3) The longitudinal study over years of the evolution of the mechanical stress and of the

potential links between this evolution and the emergence of speech production deficits

and /or swallowing diseases.

The measure of the mechanical stress will use the original

device elaborated at Gipsa-lab (Jeannin et al., 2008) (see

opposite figure). This device consists in 12 strain-gauge

sensors inserted in an exact replicate of the dental prosthesis

and distributed over the whole surface of the prosthesis. The

strain-gauge sensors have been specifically developed for this

purpose and the whole electronic setup matches the safety

regulation for an insertion in a patient’s mouth. Its insertion

within the dental prosthesis enables that it is not detectable by

the patient. Thus, its presence does not provide by itself a

perturbation of the motor task. Only the wires getting out from

the mouth corner are likely to induce a perturbation..

With this experimental study, we’d like to establish how patients adapt to the presence of the

prosthesis in their mouth, and how it can lead to the emergence of new, potentially pathological, motor

control strategies of speech production and swallowing. We make the hypothesis that this adaptation

could involve at least two different stages along which problems can emerge: (1) the elaboration in the

brain of an internal representation (a so-called internal model) of the oro-facial motor system including

the prosthesis, which memorizes the new links between motor commands, oro-sensory feedback and

accuracy of the achievement of the motor task (Wolpert et al., 1998); (2) the evolution of the motor

strategies on the basis of this new internal model. In line with studies of speech production in cochlear

implanted patients carried out by Perkell et al (2000), it can be expected that the decrease of the

accuracy of the afferent information due the presence of the prosthesis could induce a difficulty in the

maintenance of the internal models and, consequently, a decrease in the accuracy of the motor task.

The ultimate goal of the study is (1) via the establishment of a map of the contacts and their intensity

between the tongue and the prosthesis, to provide useful information for the design of prostheses and

of their insertion in the patient’ mouth, and (2) via the behavioral study of patients, to contribute to the

design of adapted training and reeducation protocols.

References

Jeannin, C., Perrier, P., Payan, Y., Dittmar, A., & Grosgogeat, B. (2008). Tongue pressure recordings during speech using

complete denture. Materials Science and Engineering C; 28, 835-841.

Perkell, J. S., Guenther, F. H., Lane, H., … & Zandipour, M. (2000). A theory of speech motor control and supporting data from

speakers with normal hearing and with profound hearing loss. Journal of Phonetics, 28(3), 233–272.

Wolpert, D., Miall, R. C., and Kawato, M. (1998). Internal models in the cerebellum. Trends in Cognitive Sciences, 2(9), 338–

347.

Skills

Taste for experimental work.

Ability to deal with patients and elderly people.

Statistical analysis of behavioral data.

Interest for cognitive sciences in general and learning issues in particular.

Salary:

According to French standards (about 1400 € net/month)

How to apply?

In order to apply, interested candidates should send from now their application (CV, transcript of

results, motivation letter and at least one reference) to:

Christophe.Jeannin@univ-lyon1.fr

Brigitte.Grosgogeat@univ-lyon1.fr

Pascal.Perrier@gipsa-lab.grenoble-inp.fr

If more information is needed, do not hesitate to contact us.

Back

Top

La mise au point de services facilitant l’accès aux contenus de la Connaissance constitue l’un des axes de recherche portés par Orange Labs Research. L'objet du postdoc est d'étudier la structuration automatique de contenus dans ce domaine (conférences, documentaires, entretiens, émissions de radio ou TV spécialisées, …).

La structuration thématique de contenus est un sujet d’étude actif depuis un certain nombre d’années dans la communauté scientifique. Le sujet a été plus largement traité dans le cadre de la segmentation thématique de journaux télévisés mais également pour des contenus de nature différente comme des réunions ou des cours magistraux (lecture topic segmentation). Dans le cas de la segmentation thématique de contenus de type 'journaux télévisés', les sujets consécutifs sont le plus souvent suffisamment différenciés pour que les techniques basées sur la cohésion lexicale permettent d'obtenir de bons résultats. Le matériau de base pour cette segmentation, à savoir la transcription automatique du contenu, est par ailleurs le plus souvent de bonne qualité, du fait des bonnes performances des systèmes de transcription qui ont été largement optimisés dans la communauté parole sur des contenus de ce type.

Les contenus étudiés dans le cadre de ce post-doc (conférences, documentaires, entretiens) ont la particularité d'être globalement monothématiques, potentiellement longs comme dans le cas des conférences, et potentiellement très spécialisés. L'adaptation de la segmentation thématique à la segmentation en sous-thème pour ce type de contenus constitue le principal objectif du travail.

Plusieurs questions scientifiques devront être abordées :

* Au-delà de la cohésion lexicale entre les mots pris individuellement, il s'agira de s'attacher à la cohésion lexicale entre les mots en contexte (dans le cas d'une conférence sur le cerveau, le terme cortex pourra apparaître à plusieurs moment alors que les contextes 'cortex visuel', 'cortex frontal' seront peut-être plus discriminants pour la structuration). La question est alors de trouver une définition appropriée du contexte.

* En amont de la structuration, comment enrichir la couverture lexicale pour le traitement automatique de documents spécialisés?

* Lors du processus de structuration, comment tirer profit d'une base de connaissance structurée, de relations sémantiques? Comment définir et exploiter des mesures d’association distributionnelles ?

* Comment proposer une organisation hiérarchique des sous-thèmes afin de proposer une structuration de type 'plan' des documents?

Les résultats du post-doc permettront d’alimenter des travaux de recherche autour d’interfaces innovantes de navigation dans les contenus de la Connaissance.

Le candidat devra être titulaire d’un doctorat en informatique, spécialisé en traitement de la parole, traitement du langage ou en recherche d’information.

Le postdoc se déroulera à Lannion pour une période de 12 mois et doit débuter en 2015. Le candidat est recruté en CDD, avec une rémunération allant de 35 à 38 keuros brut annuel. Les contraintes de recrutement en CDD imposent que le candidat n’ait pas réalisé plus d’un postdoc préalablement à la candidature.

Contact Géraldine Damnati : geraldine.damnati@orange.com

Back

Top

Post-doc position at LORIA (Nancy, France)

Automatic speech recognition: Deep Neural Network for language model

Framework of ANR project ContNomina

The technologies involved in information retrieval in large audio/video databases are often based on the analysis of large, but closed, corpora, and on machine learning techniques and statistical modeling of the written and spoken language. The effectiveness of these approaches is now widely acknowledged, but they nevertheless have major flaws, particularly for what concern proper names, that are crucial for the interpretation of the content.

In the context of diachronic data (data which change over time) new proper names appear constantly requiring dynamic updates of the lexicons and language models used by the speech recognition system.

As a result, the ANR project ContNomina (2013-2017) focuses on the problem of proper names in automatic audio processing systems by exploiting in the most efficient way the context of the processed documents. To do this, the post-doc student will address the contextualization of the recognition module through the dynamic adjustment of the language model in order to make it more accurate.

Post-doc subject

Deep Neural Network have become a key component of modern automatic speech recognition systems. The language model of our recognition system is based on a neural network learned from a large corpus of text. The problem is to estimate the probability of a new proper name depending on its context. Several tracks will be explored: adapting the language model, using a class model or studying the notion of analogy.

Our team has developed a fully automatic system for speech recognition to transcribe a radio broadcast from the corresponding audio file. The post-doc will develop a new module whose function is to integrate new proper names in the language model.

Required skills

A PhD in signal processing or in computer sciences, be familiar with the tools for automatic speech recognition, background in statistics and computer program skills (??C, object-oriented programming and Perl).

Post-doc duration: 12 months, start during the 2015 (these is some flexibility)

Localization and contacts: Loria laboratory, Speech team, Nancy, France

irina.illina@loria.frdominique.fohr@loria.fr georges.linares@univ-avignon.fr

Candidates should email a letter of application, a detailed CV with a list of publications and diploma

Back

Top

Recrutement de 2 postdocs en sciences du language/linguistique/psychologie

Institution : Laboratoire Parole et Langage (CNRS, Aix-Marseille Université)

Nombre de postes :  2

Durée :  2 ans (voir ci-dessous)

Date limite de reception des candidatures : 1 juin 2015

Le Laboratoire Parole et Langage, Unité Mixte de Recherche entre le CNRS et Aix-Marseille Université, associé au LABEX « Brain and Language Research Institute (BRLI) » ouvre un appel à candidature pour 2 postes « post-doctoraux », financés par la fondation  A*MIDEX .  Le projet vise à étudier la relation entre des variables sociales et les représentations linguistiques  et cherche à développer un modèle cognitif qui prenne en compte les effets d?identité sociale sur la perception et la production de la parole. Le projet combinera l?expérimentation (production, perception, potentiels évoqués), l?analyse des corpus et les modèles computationnels. Le candidat fera partie intégrale de l?équipe interdisciplinaire du BLRI, qui implique des chercheurs en linguistique, psychologie, neurosciences, et informatique.

Le projet se focalise sur les aspects prosodiques et intentionnels de la communication, de telle sorte qu?une expérience dans ces domaines sera un atout. Par ailleurs, les candidats qui auront des compétences dans l?un ou plusieurs domaines mentionnés ci-dessous seront prioritaires :

-        Sociolinguistique (quantitative)

-        Approche stochastique/probabiliste de  la théorie linguistique

-        Psychologie/psycholinguistique de mémoire (surtout encodage et recouvrement)

-        Phonologie de laboratoire

-        Linguistique expérimentale

Le poste sera financé pour une durée de 24 mois au maximum en fonction de la date du début du contrat. Le candidat aura obtenu sa thèse avant le début du contrat. La date du début du contrat est flexible mais le poste devra être pourvu avant le 1 octobre 2015.

Les demandes peuvent être soumises en français ou en anglais, par courriel à Jessica Bouanane (responsable administrative) E-mail :  jessica.bouanane@univ-amu.fr.

Pour plus de renseignements veuillez contacter James Sneed German (investigateur principal); E-Mail : james.german@lpl-aix.fr.

Back

Top

Master’s, PhD, and Postdoc Openings at the MuSAE Lab in Montreal, Canada

The Multimedia Signal Analysis and Enhancement (MuSAE) Lab has openings at the graduate and post-graduate levels to work in the area of non-intrusive objective speech quality measurement of enhanced speech. Particular interest is placed on microphone array processed speech. The recruited researcher(s) will work closely with national and international industry (e.g., Google) and academic partners.

Candidates should have strong knowledge in at least two of the following topics: speech signal processing, single- and multi-channel speech enhancement, subjective/objective speech quality assessment, psychoacoustics, and pattern recognition (eg, deep learning). Hands-on experience with commercial microphone arrays, as well as with speech data collection in noisy and reverberant environments will be considered as great assets. Interested candidates should send their transcripts (unofficial copies), up-to-date CV's, and relevant publications to Dr. Tiago Falk (falk [at] emt [dot] inrs [dot] ca). Please note that only short-listed candidates will be contacted for an interview. Moreover, postdoctoral candidates will be expected to have a strong publication record in reputable journals and conference proceedings.

The MuSAE Lab is located at the Institut National de la Recherche Scientifique (INRS-EMT), University of Quebec, in Montreal, Canada – a bilingual (English/French) city renowned for its great pool of ethnic diversity and its vibrant cosmopolitan ambiance. INRS has consistently ranked as a top-3 Canadian university in the category ‘Research Intensity and publications’ and houses several state-of-the-art research facilities. More detail about the MuSAE Lab and its research team can be found at: http://musaelab.ca.

Back

Top

PhD Research Fellow

Speech Production and Swallowing with Full Dental Prosthesis

(May 28th, 2015)

Context

A position as PhD Research fellow is available in the PCMD (« Perception, Contrôle, Multimodalité et

Dynamiques de la parole ») research team in that GIPSA-Lab. This PhD is part of a research project

involving researchers from two different labs, Gipsa-lab in Grenoble and the LMI in Lyon. It is funded

by the Région Rhône-Alpes for 3 years from next fall, in the context of the research program ARC2

focusing on Quality of Life in the Elderly.

Scientific goals

The general objective of the thesis is to thoroughly study in patients the mechanical interactions

between a full dental prosthesis and the tongue during speech production and swallowing, and to

understand how it influences the control of these basic orofacial motor tasks and its evolution over

time. The work will be organized around three main tasks:

(1) The characterization of the mechanical stress exerted by the tongue on the dental

prosthesis, with a special focus on its amplitude, its spatial distribution and its variation

across patients.

(2) The study of the variation of the mechanical stress during the first weeks following the

insertion of the dental prosthesis in the patient’s mouth, during which a motor adaptation

takes place.

(3) The longitudinal study over years of the evolution of the mechanical stress and of the

potential links between this evolution and the emergence of speech production deficits

and /or swallowing diseases.

The measure of the mechanical stress will use the original

device elaborated at Gipsa-lab (Jeannin et al., 2008) (see

opposite figure). This device consists in 12 strain-gauge

sensors inserted in an exact replicate of the dental prosthesis

and distributed over the whole surface of the prosthesis. The

strain-gauge sensors have been specifically developed for this

purpose and the whole electronic setup matches the safety

regulation for an insertion in a patient’s mouth. Its insertion

within the dental prosthesis enables that it is not detectable by

the patient. Thus, its presence does not provide by itself a

perturbation of the motor task. Only the wires getting out from

the mouth corner are likely to induce a perturbation..

With this experimental study, we’d like to establish how patients adapt to the presence of the

prosthesis in their mouth, and how it can lead to the emergence of new, potentially pathological, motor

control strategies of speech production and swallowing. We make the hypothesis that this adaptation

could involve at least two different stages along which problems can emerge: (1) the elaboration in the

brain of an internal representation (a so-called internal model) of the oro-facial motor system including

the prosthesis, which memorizes the new links between motor commands, oro-sensory feedback and

accuracy of the achievement of the motor task (Wolpert et al., 1998); (2) the evolution of the motor

strategies on the basis of this new internal model. In line with studies of speech production in cochlear

implanted patients carried out by Perkell et al (2000), it can be expected that the decrease of the

accuracy of the afferent information due the presence of the prosthesis could induce a difficulty in the

maintenance of the internal models and, consequently, a decrease in the accuracy of the motor task.

The ultimate goal of the study is (1) via the establishment of a map of the contacts and their intensity

between the tongue and the prosthesis, to provide useful information for the design of prostheses and

of their insertion in the patient’ mouth, and (2) via the behavioral study of patients, to contribute to the

design of adapted training and reeducation protocols.

References

Jeannin, C., Perrier, P., Payan, Y., Dittmar, A., & Grosgogeat, B. (2008). Tongue pressure recordings during speech using

complete denture. Materials Science and Engineering C; 28, 835-841.

Perkell, J. S., Guenther, F. H., Lane, H., … & Zandipour, M. (2000). A theory of speech motor control and supporting data from

speakers with normal hearing and with profound hearing loss. Journal of Phonetics, 28(3), 233–272.

Wolpert, D., Miall, R. C., and Kawato, M. (1998). Internal models in the cerebellum. Trends in Cognitive Sciences, 2(9), 338–

347.

Skills

Taste for experimental work.

Ability to deal with patients and elderly people.

Statistical analysis of behavioral data.

Interest for cognitive sciences in general and learning issues in particular.

Salary:

According to French standards (about 1400 € net/month)

How to apply?

In order to apply, interested candidates should send from now their application (CV, transcript of

results, motivation letter and at least one reference) to:

Christophe.Jeannin@univ-lyon1.fr

Brigitte.Grosgogeat@univ-lyon1.fr

Pascal.Perrier@gipsa-lab.grenoble-inp.fr

If more information is needed, do not hesitate to contact us.

Back

Top

INTITULE DU POSTE : Ingénieur en techniques expérimentales

Catégorie : Ingénieur d?étude

Service/Composante : UFR Arts, Lettres, Langues et Sciences Humaines (ALLSH)

Localisation du poste : Maison de la recherche

BAP : C

Emploi type : Ingénieur en techniques expérimentales

Date de vacance du poste : 01/09/2015

Quotité : 100 %

Date limite de candidature : 26 juin 2015

CONTEXTE

La maison de la recherche de l?UFR ALLSH se situe dans un bâtiment neuf de 4 500 m2 destiné à

l?accueil d?unités de recherche en Arts, Lettres, Langues et Sciences Humaines (9 Equipes d?Accueil :

PSYCLE, LPS, LPCLS, LERMA, CIELAM, CAER, IHP, ECHANGES, LESA et deux Unités Mixtes de

Recherche : IRASIA, CEPERC), des Écoles Doctorales en Lettres Sciences Humaines (ED 354&356) et

du service des Presses Universitaires. Les locaux de recherche ont été organisés pour être

spatialement cohérents, fonctionnels et équipés selon les besoins et spécificités des Unités, pour un

développement optimal et une visibilité accrue de la recherche sur ce secteur scientifique.

Le poste d?ingénieur en techniques expérimentales s?insère dans un ensemble de services mutualisés

et créés spécifiquement à l?occasion de l?ouverture de la Maison de la Recherche.

L?Ingénieur en techniques expérimentales en soutien scientifique et technique à la recherche au sein

des laboratoires sus mentionnés viendra en appui au fonctionnement des plateformes scientifiques

de la Maison de la Recherche : les salles d?expérimentation et d?observation en psychologie et le

studio d?enregistrement mais également aux équipements de recherche mobiles.

Cet appui permettra d?impulser de nouveaux programmes innovants de recherche.

MISSIONS ET ACTIVITES PRINCIPALES :

Procéder aux montages, réglages, et essais d?appareils ou de montages expérimentaux.

Mettre au point les dispositifs expérimentaux ; définir, développer, tester et formaliser les

protocoles.

Concevoir les adaptations et les améliorations de tout ou partie d?un dispositif expérimental.

Coordonner l?exploitation du dispositif.

Formation à la technique et à l?utilisation des dispositifs expérimentaux ; conseil aux utilisateurs

pour leur mise en oeuvre dans le respect des normes d?utilisation.

Rédiger les documents de spécifications techniques, de conception et de réalisation et les manuels

utilisateurs associés aux dispositifs expérimentaux.

Assurer le bon déroulement de l?expérimentation.

Participer à la valorisation des technologies des plateformes.

Suivre les évolutions des techniques expérimentales et se former pour les mettre en oeuvre.

Organiser et contrôler les interventions de maintenance préventive et les interventions de

dépannage.

Ce poste est ouvert en externe aux titulaires de la fonction publique de catégorie A (IGE) et aux

agents non titulaires détenteur minimum d?un Bac + 3.

Le dossier de candidature (CV et LM avec lorsque cela est possible, le dernier entretien

professionnel) sera adressé par voie électronique à :

christine.djambazian@univ-amu.fr

et

natacha.mauric@univ-amu.fr.

Back

Top

UPCOMING DEADLINE - JUNE 29TH 2015 - 6PM ITALIAN TIME
PROSSIMA SCADENZA - 29 GIUGNO 2015 - 18:00 ORA ITALIANA

We are pleased to inform you about the IMT Institute for Advanced Studies Lucca's call for applications for PhD positions in the following specialized curricula:
Si informa che è aperto il nuovo bando di selezione per l'accesso al Programma di Dottorato di IMT Alti Studi Lucca per i seguenti curricula di specializzazione:

• Analysis and Management of Cultural Heritage
• Computer Science and Systems Engineering
• Economics, Management and Data Science

• No tuition fees, free room and access to IMT Canteen
  Esenzione dalle tasse d'iscrizione, alloggio gratuito, accesso gratuito alla mensa IMT
• A grant of €13,630 gross/year
  Importo lordo della borsa: €13.630/anno

Candidates can apply if they obtain their (minimum) 4-year undergraduate degree NO LATER than October 30^th, 2015
Possono fare domanda coloro che conseguano il titolo di accesso entro e NON OLTRE 30 ottobre 2015

DEADLINE: June 29^th, 2015 at 6:00 pm CEST
MORE INFORMATION: http://phd.imtlucca.it

IMT Lucca (www.imtlucca.it) is a research university within the Italian public higher education system. IMT's research mission is characterized by the analysis and management of a plurality of economic, social, technological and cultural systems. Faculty carry out cutting edge methodological research to generate knowledge, to design projects to apply knowledge to resolve economic, industrial, social and cultural problems, and to transmit knowledge through highly-specialized graduate courses and seminars.
IMT Lucca (www.imtlucca.it) è un istituto universitario pubblico di ricerca e alta formazione dottorale a ordinamento speciale, focalizzato sull'analisi e gestione dei sistemi economici, sociali, tecnologici e culturali. I docenti e i ricercatori dell'Istituto svolgono attività di ricerca metodologica di alto livello scientifico per generare conoscenza, per applicare la conoscenza alla risoluzione di problemi d'interesse economico, industriale, sociale e culturale attraverso la realizzazione di progetti, e per trasmettere la conoscenza attraverso l'attività didattica.

Follow us on Facebook, Twitter, YouTube, LinkedIn.

Please note that this account is used for outgoing e-mails only, and therefore your reply message will not be read. For inquiries regarding the 2015/16 IMT PhD call, please write to phdapplications@imtlucca.it.
If you no longer wish to receive emails from IMT Lucca, please send an email with the word UNSUBSCRIBE in the subject line to commev@imtlucca.it.

Ufficio Offerta didattica, Dottorato e Servizi agli Studenti
Graduate Education and Student Services Office
IMT Alti Studi Lucca
Piazza San Ponziano 6, 55100 Lucca - Italia
Tel: +39 0583 4326 745, fax: +39 0583 4326 565
www.imtlucca.it

Back

Top

A one-year post-doc position open for this fall at Clemson University.

 Candidate should have previous research experience and skills in one or more of virtual environments, 3DUI, experimental design and evaluation, intelligent virtual agents, and stereoscopic computer graphics.

The virtual environments group at Clemson is a dynamic research group with approximately 25 faculty, post-doc and student participants.

Applications should be emailed directly to

and should include a C.V., statement of research interests, and the names of three references.

Back

Top

Le LPP recrute un poste d'ingénieur d'étude CNRS en déploiement et développement d'application par le biais d'un concours externe

Back

Top

We have a 3 year fully-funded PhD studentship in the use of Spoken Language Dialogue Systems in Assistive Technology. Full details are at

Back

Top