Papers by Laurence Devillers
Springer eBooks, 2022
Over the past decade, eye movement has been widely looked into for describing and analyzing sever... more Over the past decade, eye movement has been widely looked into for describing and analyzing several cognitive processes and especially for human-document interaction, such as estimating reading ability and document understanding. Most of the existing applications have been done for silent reading but we propose to explore reading aloud interaction through a powerful measurement named the "eye-voice span" which measures the distance between the eyes and the voice. In this paper we present an open-source platform named "Eye got it" and the underlying algorithms that can be used for processing eye-tracking and voice data in order to compute automatically the eye-voice span.
Symbiotic Interaction, 2017
We demonstrate that the current Child-Computer Interaction paradigm is not potentiating human dev... more We demonstrate that the current Child-Computer Interaction paradigm is not potentiating human development to its fullest. It is linked to several physical and mental health problems and appears not to be maximizing children's cognitive development and performance. To potentiate children's physical and mental health (including cognitive development and performance) we conceived a new approach to human development and evolution. This approach proposes a particular synergy between the developing human body, computing machines and natural environments. It emphasizes that children should be encouraged to interact with challenging physical environments offering multiple possibilities for sensory stimulation and increasing physical and mental stress to the organism. We created and tested a new set of robotic devices to operationalize our approach -Biosymtic (Biosymbiotic Robotic) devices.
International Cconference on Multimodal Interaction
Robot-directed speech refers to speech to a robotic device (speakers, computers, etc.). Studies h... more Robot-directed speech refers to speech to a robotic device (speakers, computers, etc.). Studies have investigated the phonetic and linguistic properties of this type of speech and shown that, humans tend to change their pitch when talking to a robot vs to a human. Parallelly, it has shown that the anthropomorphism of the devices afects the social aspect of interaction. However, none have investigated the efect of the device's human-likeliness on linguistic realizations. This study proposes to fll this gap by comparing the efect of anthropomorphism in speech directed at a speaker vs a humanoid robot vs a human by analyzing the F0 values and range in the three conditions, and how these parameters change throughout the conversation. The data from 52 native speakers of French show that robot-directed speech shares several pitch tendencies with speaker-directed speech, which in its turn is situated between human-and robot-directed speech. CCS CONCEPTS • Human-centered computing → Human computer interaction (HCI); Interaction design.
Proceedings of the 2022 ACM International Joint Conference on Pervasive and Ubiquitous Computing
The circulation of knowledge is an important function for enriching our social life. Knowledge ci... more The circulation of knowledge is an important function for enriching our social life. Knowledge circulation, which has traditionally been conducted manually, is undergoing a transformation in the age of artificial intelligence. In this paper, we describe our project called "Learning Cyclotron (LeCycl), " which was initiated to accelerate the circulation of knowledge for building a knowledge ecosystem based on artificial intelligence technology. The three functions of sensing, mastering, and transferring knowledge are effectively operated through the power of AI-empowered digital nudging strategies. We outline what has been accomplished to date and summarize future directions for the ultimate goal of LeCycl.
Proceedings of the Seventh …, 2010
In this paper we describe a corpus set together from two sub-corpora. The CINEMO corpus contains ... more In this paper we describe a corpus set together from two sub-corpora. The CINEMO corpus contains acted emotional expression obtained by playing dubbing exercises. This new protocol is a way to collect mood-induced data in large amount which show several complex and shaded ...
ISLE Workshop, 2002
For the AMITIÉS multilingual human-computer dialogue project [1], we have developed new methods f... more For the AMITIÉS multilingual human-computer dialogue project [1], we have developed new methods for the manual annotation of spoken dialogue transcriptions from European financial call centers on multiple levels. We have modified the DAMSL schema [2] to create a dialogue act taxonomy appropriate to the functions of call center dialogues. We use a domain-independent framework populated with domain-specific lists to capture the semantics of spoken dialogues. Our new flexible, platform-independent Java annotation tool, called XDMLTool, takes plain-text dialogue files as input, and yields annotated files in the widely used XML format. To date, XDMLTool has been used to annotate several hundred call-center dialogues in France, the UK and the US. We present definitions of each tag as well as examples in English and French. These annotation methods are developed for an experimental system that automates financial call centers in Europe. The multi-level annotation scheme has been used to develop a prototype triaging application for financial services.
The HUMAINE Emotion Database
Proceedings 2nd European Language Resources and Technologies Forum: Language Resources of the future–the future of Language Resources,(Barcelona, Spain), FLaReNet, Feb 1, 2010
Recognition of emotion in speech has recently matured to one of the key disciplines in speech ana... more Recognition of emotion in speech has recently matured to one of the key disciplines in speech analysis serving next generation human-machine, human-robot communication, and media retrieval systems. Numerous studies have been seen in the last decade trying to improve on features and classifiers. However, in comparison to related speech processing tasks such as Automatic Speech and Speaker Recognition, practically no standardized corpora and test-conditions exist to compare performances under exactly the same ...
Proc. 3rd ELR0A Internat. Workshop on EMOTION (satellite of LREC): Corpora for Research on Emotion and Affect, Valetta, 2010
In this paper we suggest feature selection and Principal Component Analysis as a way to analyze a... more In this paper we suggest feature selection and Principal Component Analysis as a way to analyze and compare corpora of emotional speech. To this end, a fast improvement of the Sequential Forward Floating Search algorithm is introduced, and subsequently extensive tests are run on a selection of French emotional language resources well suited for a first impression on general applicability. Tools for comparing feature-sets are developed to be able to evaluate the results of feature selection in order to obtain ...
MEDIA: évaluation de la compréhension dans les systèmes de dialogue
... Media : évaluation de la compréhension dans les systèmes de dialogue. Hélène Bonneau-Maynard ... more ... Media : évaluation de la compréhension dans les systèmes de dialogue. Hélène Bonneau-Maynard 1 , Alexandre Denis 2 , Frédéric Béchet 3 , Laurence Devillers 1 , Fabrice Lefèvre 3 , Matthieu Quignard 2 , Sophie Rosset 1 , Jeanne Villaneau 4. (2008). ...
Annotation of Affective Interaction in Real-life Dialogs Collected in a Call-center
The Workshop Programme
Annotation of Affective Interaction in Real-life Dialogs Collected in a Call-center Christophe Va... more Annotation of Affective Interaction in Real-life Dialogs Collected in a Call-center Christophe Vaudable, Nicolas Rollet, Laurence Devillers LIMSI-CNRS Orsay, France E-mail: christophe. vaudable@ limsi. fr, devil@ limsi. fr Abstract The originality of this paper is to propose an ...
Fiction database for emotion detection in abnormal situations
Proc. of ICSLP, Jeju, 2004
The present research focuses on the acquisition and an-notation of vocal resources for emotion de... more The present research focuses on the acquisition and an-notation of vocal resources for emotion detection. We are interested in detecting emotions occurring in abnor-mal situations and particularly in detecting fear. The present study considers a preliminary database of ...
Editorial: Special issue of computer speech and language on affective speech in real-life interactions
Within the speech and language processing communitites there is considerable and growing interest... more Within the speech and language processing communitites there is considerable and growing interest in issues related to emotion and affect in speech (see e.g., the Humaine workshop held immediately prior to and as a satellite of this conference). However, the terms "emotion" and "affect" are often used almost interchangeably. The goal of this panel discussion will be first to define and differentiate the two terms, as they relate to speech processing, and then to specify the different needs and requirements of research and technology development for each. All panelists have experience in these fields of speech and language processing, and will be able to call on their own experience as well as that gained from discussions in the half-day workshop preceeding LREC "Corpora for Research on Emotion And Affect". We look forward to lively contributions from the floor and hope that the discussion will allow us to establish a common ground between the various disciplines engaged in collecting related corpora so that a better understanding of the needs of each community may be achieved.
Patterns, Prototypes, Performance
Patterns, Prototypes, Performance A. Batliner1, D. Seppi2, B. Schuller3, S. Steidl1, T. Vogt4, J.... more Patterns, Prototypes, Performance A. Batliner1, D. Seppi2, B. Schuller3, S. Steidl1, T. Vogt4, J. Wagner4, L. Devillers5, L. Vidrascu5, N. Amir6, and V. Aharonson7 1 Chair of Pattern Recognition (LME), University Erlangen-Nuremberg, Germany 2 Fondazione Bruno Kessler (FBK)–irst, Trento, Italy 3 Institute for Human-Machine Communication, Technische Universität München (TUM), Germany 4 Multimedia Concepts and their Applications, University of Augsburg (UA), Germany 5 Spoken Language Processing Group (LIMSI-CNRS), Orsay Cedex, France 6 ...
Traditionally, it has been assumed that pitch is the most important prosodic feature for the mark... more Traditionally, it has been assumed that pitch is the most important prosodic feature for the marking of prominence, and of other phenomena such as the marking of boundaries or emotions. This role has been put into question by recent studies. As nowadays larger databases are always being processed automatically, it is not clear up to what extent the possibly lower relevance of pitch can be attributed to extraction errors or to other factors. We present some ideas as for a phenomenological difference between pitch and duration, and compare the performance of automatically extracted F0 values and of manually corrected F0 values for the automatic recognition of prominence and emotion in spontaneous speech (children giving commands to a pet robot). The difference in classification performance between corrected and automatically extracted pitch features turns out to be consistent but not very pronounced.
Classification performance of emotional user states found in realistic, spontaneous speech is not... more Classification performance of emotional user states found in realistic, spontaneous speech is not very high, compared to the performance reported for acted speech in the literature. This might be partly due to the difficulty of providing reliable annotations, partly due to suboptimal feature vectors used for classification, and partly due to the difficulty of the task. In this paper, we present a cooperation between several sites, using a thoroughly processed emotional database. For the four-class problem motherese/neutral/emphatic/angry, we first report classification performance computed independently at each site. Then we show that by using all the best features from each site in a combined classification, and by combining classifier outputs within the ROVER framework, classification results can be improved; all feature types and features from all sites contributed. Združevanje sil za boljše samodejno razvrščanječustvenih stanj uporabnika: Uspešnost samodejnega razvršžanjačustvenih stanj uporabnika, ki jih najdemo v realističnem, spontanem govoru, je v primerjavi s kakovostjo, ki jo v literaturi navajajo za igrani govor, precej nižja. To je lahko delno posledica težav pri zagotavljanju zanesljive anotacije, delno posledica uporabe podoptimalnih vektorjev značilk pri razvršžanju, delno pa posledica težavnosti te naloge. V prispevku predstavljamo sodelovanje med različnimi ustanovami na temeljito obdelani bazi podatkov. Zaštiristopenjski problem govor otroku/nevtralno/poudarjeno/jezno najprej navedemo kakovost razvršžanja, kot so jo izračunali neodvisno na vsaki od sodelujočih ustanov. Nato poka-žemo, da lahko izboljšamo rezultate razvršžanja z uporabo najboljših značilk vsake izmed ustanov in z združevanjem rezultatov razvršževalnikov znotraj ogrodja ROVER.
Résumé: This paper presents and reports on the progress of the EVALDA/MEDIA project, focusing on ... more Résumé: This paper presents and reports on the progress of the EVALDA/MEDIA project, focusing on the recording protocol of the reference dialogue corpus. The aim of this project is to define and test an evaluation methodology that assess and diagnose the contextsensitive understanding capability of spoken language dialogue systems. Systems from both academic organizations (CLIPS, IRIT, LIA, LIMSI, LORIA, VALORIA) and industrial sites (FRANCE TELECOM R et D, TELIP) will be evaluated. ELDA is the coordinator of the ...
In: Proceedings of the Fourth International Conference On Language Resources and Evaluation (LREC), Lisbon, May 1, 2004
This paper will present and report on the progress of the EVALDA/MEDIA project, focusing on the r... more This paper will present and report on the progress of the EVALDA/MEDIA project, focusing on the recording and annotating protocol of the reference dialogue corpus. The aim of this project is to design and test an evaluation methodology to compare and diagnose the context-dependent and independent understanding capability of spoken language dialogue systems. Systems from both academic organisations (IRIT, LIA, LIMSI, LORIA, VALORIA, CLIPS) and industrial sites (FRANCE TELECOM R&D, TELIP) will be ...
Affective Databases, Annotations, Tools and Languages-The HUMAINE Database: Addressing the Collection and Annotation of Naturalistic and Induced Emotional Data
Lecture Notes in Computer Science, 2007
Uploads
Papers by Laurence Devillers