<html> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"> </head> <body text="#000000" bgcolor="#FFFFFF"> <div style="text-align: center;" class="">PhD fellowship</div> <div style="text-align: center;" class=""> </div> <div style="text-align: center;" class=""> Computational model of conversational behaviors integrating behavioral and physiological data from human-human and human-machine interactions </div> <div style="text-align: center;" class=""> </div> <div style="text-align: center;" class=""> Laboratoire d’Informatique et des Systèmes (LIS) et Laboratoire Parole et Langage (LPL) <del cite="mailto:ochsm" datetime="2018-03-09T13:30"></del> Aix-Marseille Université & CNRS</div> <div class=""> </div> Keywords: conversational speech, multimodal data analysis, neurophysiological data, machine learning, artificial agents. The PhD project is part of an A*MIDEX project PhysSocial that aims at a better understanding of the specificities of social interactions by comparing relationships between behavior and neurophysiology in human‐human and human‐robot discussion. The goal of the PhD is to<ins cite="mailto:PREVOT%20Laurent" datetime="2018-03-03T10:08"> </ins>analyze the multimodal signals (speech, eyes direction, physiological, and neurophysiologic signals) from conversational activity using signal processing and machine learning methodologies in order to compare the human-human and human-robot interactions. The PhD is organized around 3 main tasks: <ul> <li> <span style="font-style: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-size-adjust: none; font-stretch: normal; font-feature-settings: normal; font-language-override: normal; font-kerning: auto; font-synthesis: weight style; font-variant: normal;">Multimodal data preprocessing: in a first step, the objective is to process the row data (speech, transcribed speech, eyes tracking, physiological and neurophysiological signals) corresponding to human-human and human-robot conversation in order to extract time series corresponding to behavioral features, as well as cognitive events derived from local activity in well-defined brain areas involved in language and social cognition</li> <li><span style="font-style: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-size-adjust: none; font-stretch: normal; font-feature-settings: normal; font-language-override: normal; font-kerning: auto; font-synthesis: weight style; font-variant: normal;">Machine learning of causal relations: in a second step, time series will be used by statistical learning to identify causal relations between behavioral and physiological features and cognitive events extracted from neurophysiological recording with fMRI. From a learning point of view, one challenge in this project is the high-dimensional data. We address this issue with a focus on the features representation and selection problems.</li> <li><span style="font-style: normal; font-weight: normal; font-size: 7pt; line-height: normal; font-size-adjust: none; font-stretch: normal; font-feature-settings: normal; font-language-override: normal; font-kerning: auto; font-synthesis: weight style; font-variant: normal;">Computational modeling and implementation in a humanoid robot: the last step consists in integrating the knowledge extracted from the data sets into an existing platform (Furhat talking head) in order to generate the appropriate behavior (speech and eyes behavior) of the artificial agent during an interaction with an user. The model will be evaluated through a experimentation that will be conducted in collaboration with the other partners of the project. </li> </ul> The PhD candidate should have a master's degree completed in Computer Science, Applied Mathematics, Signal or Natural Language Processing (with solid background in machine learning). The candidate should have a strong background in machine learning and signal processing with a focus on multimodality. Some complementary previous experience would be appreciated in the following topics: • Multimodal data processing • Data science applied to language data • Dialogue systems The PhD is fully funded during 3 years as part of the A*MIDEX interdisciplinary project PhysSocial, including personalized training, travel expenses, and conferences attendance. French language is not required. Aix Marseille University (<a href="http://www.univ-amu.fr/en" target="_blank">http://www.univ-amu.fr/en</a>), the largest French University, is ideally located on the Mediterranean coast, and only 1h30 away from the Alps. The application files consists of the following documents: - A detailed curriculum, - A description of the academic background and copy of academic records and most recent diploma, - A cover letter describing why the applicant wishes to participate in this project, a justification of the inter-disciplinary and international aspects of her/his research, his/her training project and career Plan including these dimensions, and his/her research’s adequacy with the proposed topics - 2 recommendation letters (including one from the master or equivalent diploma supervisor) The application files should be sent to : Laurent Prévot: <a href="mailto:laurent.prevot@univ-amu.fr">laurent.prevot@univ-amu.fr</a> and Magalie Ochs: <a href="mailto:magalie.ochs@lis-lab.fr">magalie.ochs@lis-lab.fr</a> Please send your intention to apply to this PhD position with your CV and the copy of the academic files before July 15th For any question, contact : Laurent Prévot: <a href="mailto:laurent.prevot@univ-amu.fr">laurent.prevot@univ-amu.fr</a> <a href="http://www.lpl-aix.fr/person/*prevot*" target="_blank">www.lpl-aix.fr/person/*prevot*</a> Magalie Ochs: <a href="mailto:magalie.ochs@lis-lab.fr">magalie.ochs@lis-lab.fr</a> <a href="http://www.lsis.org/ochsm/">http://www.lsis.org/ochsm/</a> </body> </html>