Errors by Humans and Machines in multimedia, multimodal and multilingual data processing – ERRARE 2015
The Research Institute for Artificial Intelligence “Mihai Draganescu” (ICIA) of the Romanian Academy in collaboration with IMMI-CNRS, LIMSI-CNRS and the LAbEx EFL organizes the second edition of the “Errare” workshop, in September 12-13, 2015, as a satellite event of Interspeech 2015.
The workshop will be organized around the topic of errors produced and processed by humans and machines in multimedia, multimodal and multilingual data with a particular focus on spoken language. It distinguishes itself from other conferences addressing these issues by providing a forum for dialogue and exchange between researchers working in linguistics, including psycho- and neurolinguistics, on the one hand, and researchers in computer science, machine learning and multimedia speech and language processing, on the other hand. For this interdisciplinary workshop, we would like to gather these different communities around the issues of variation, ambiguity and errors in speech and language. The purpose of this workshop is to share interdisciplinary expertise on a heterogeneous phenomenon referred to as “variation” and “ambiguity” in some domains and as “errors” in others. Researchers are invited to share their thoughts and observations through case studies run in the context of various initiatives.
A large panel of research areas shares a common object of study: human language. These areas encompass historically well-established research communities: classical humanities and social sciences (phonetics, phonology, psycholinguistics, etc.), and more recent domains of the sciences (brain and computer science). Research objectives include analyzing, modeling, understanding and theorizing the human processing of speech variation. For linguists and psycholinguists variation in speech involves some matching process between variable surface forms and stable underlying forms: in such a framework errors may naturally arise as mismatches occurring at the interface of surface and underlying representations. Yet by which mechanisms errors may arise and how to interpret the patterning of errors within theoretical models of speech production and perception has been a matter of controversy. Speech error research in recent years has particularly highlighted the fuzzy boundary between the concepts of 'variability', ambiguity' and 'error'. Research activities most often include corpora consisting of various types of recorded speech from controlled (laboratory) speech to large scale data. Such corpora may be a result of a variety of capturing techniques from standard audio recordings to multi-sensor capturing of either articulation gestures or brain activities. Errors can also be envisioned as a result of noisy data capturing conditions.
Sharing experience with errors, variation and ambiguity is expected to produce beneficial insights for the different communities:
• Concerning humanities, variation and ambiguity are central to the different branches of linguistics. Furthermore, human production and perception errors challenge the existing language acquisition, production and perception models.
• For automatic speech and language processing, residual errors indicate regions which escape current modeling capacities. In-depth analyses in collaboration with linguists, psycholinguists and speech scientists may contribute to a better understanding of these phenomena and to the proposal of innovative strategies.
• Brain sciences, a recent rapidly evolving research area, open new opportunities and the study of errors can contribute to reveal the hidden organization of the brain.
------------------ TOPICS ------------------
We invite contributions focusing on errors produced by humans and/or machines from (but not limited to) the following areas:
• Cognition and brain studies related to errors in speech • Speech production (e.g. slips of the tongue...) • Speech perception • First and second language acquisition • Bilingualism and code switching • Voice pathologies / clinical phonetics
• Prosody • Natural language processing • Corpus linguistics • Automatic speech processing
• Speech and multimodality • Speech and language translation • Spoken Interaction • Information retrieval • Evaluation methods
“Errare 2015” will welcome about 80 participants, with both invited and submitted papers.
--------------- Important dates: ---------------
27 April 2015: submission deadline 15 June 2015 : notifications of acceptance 29 June 2015: final papers Workshop dates : 12-13 September 2015
--------------------- Organizing committee: ---------------------
Ioana Vasilescu (LIMSI-CNRS) Gilles Adda (IMMI-LIMSI) Joseph Mariani (IMMI-LIMSI) Verginica Mititelu (ICIA, Romanian Academy) Dan Tufis (ICIA, Romanian Academy) Maria Candea (Un iversity Paris 3) Ioana Chitoran (University Paris 7) Sophie Rosset (LIMSI-CNRS) Guillaume Wisniewski (LIMSI-CNRS) Laurence Devillers (University Paris 4/LIMSI)
------------------ Program committee: ------------------
Gilles Adda (IMMI-LIMSI) Martine Adda-Decker (University Paris 3/LIMSI) Tiberiu Boros (ICIA, Romanian Academy) Maria Candea (University Paris 3) Ioana Chitoran (University Paris 7) Laurence Devillers (University Paris 4/LIMSI) Mirjam Ernestus (Radboud University & Max Planck Institute for Psycholinguistics) Julia Hirschberg (Columbia University) Lori Lamel (LIMSI-CNRS) Mark Liberman (University of Pennsylvania) Joseph Mariani (IMMI-LIMSI) Verginica Mititelu (ICIA, Romanian Academy) Bernd T. Meyer (Carl von Ossietzky Universität Oldenburg) Marianne Pouplier (Institut für Phonetik und Sprachverarbeitung Munchen) Sophie Rosset (LIMSI-CNRS) Dan Tufis (ICIA, Romanian Academy) Ioana Vasilescu (LIMSI-CNRS) Guillaume Wisniewski (LIMSI-CNRS)
Scientific committee: To be announced soon.
------- Contact: -------
Ioana Vasilescu ioana at limsi.fr Verginica Mititelu vergi at racai.ro Gilles Adda gadda at limsi.fr Joseph Mariani joseph.mariani at limsi.fr