offre de post-doctorat‏

Voici une annonce de post-doctorat, adressée par Jean-Philippe Cointet. N’hésitez pas à la diffuser largement.

(attention : deadline le 31 mars…)

Call for JSPS Postdoctoral Fellows
Development of novel algorithms for post-disaster surveillance from massive social network data

The Japan Society for the Promotion of Science (JSPS) runs an excellent scheme for inviting international researchers to come and work at Japanese universities and research institutes for 12 to 24 months. The fellowships have offered quite generous terms that include monthly salary, round trip air ticket, housing subsidy, settling in allowance, accident and sickness coverage, local travel grant, and a grant in aid for research expenses. For conditions, application deadlines and up-to-date details please see http://www.jsps.go.jp/english/e-fellow/index.html .

NII ( http://www.nii.ac.jp/en ) and LATTICE/CNRS ( http://www.lattice.cnrs.fr/ ) are currently collaborating on a project aiming at providing novel algorithms for post-disaster surveillance from massive social network data. In this context, we would like to host applicants willing to work in this area and to be based at the Japanese-French Laboratory for Informatics (JFLI) in Tokyo for one or two years. Note that applicants from France who apply through CNRS should be either French nationals or French residents.

Please send expressions of interest on or before March, 31 by sending a CV, a cover letter and one or two relevant publications to Nigel Collier < collier@nii.ac.jp > and Thierry Poibeau < thierry.poibeau@ens.fr >. Applications are administered by CNRS with a final deadline of 16 April 2012 (see https://dri-dae.cnrs-dir.fr/spip.php?article1242 and http://jsps.unistra.fr/les-bourses-de-la-jsps/ ).

Background

Recent public health disasters such as the 2010 Haiti earthquake, the 2011 Tohoku earthquake/tsunami and the Fukushima nuclear reactor leakage have dramatically shown the human costs that societies face at times of large-scale crisis. In a very short space of time governments must decide on how to allocate scarce national assets to mitigate the disaster and alleviate suffering of victims. At the same time communities under pressure have shown great resilience and willingness to help each other. Putting together those in need with those capable of providing resources is a key objective in future disaster mitigation.

Traditional sensor networks on which planners rely such as general practitioner networks and radiation monitors are expensive to maintain and may offer insufficient coverage during times of peak requirement. For this reason we propose to explore the use of non-traditional sources of information to help governments, NPOs and international agencies identify areas of need during large scale disasters. Social media data in the form of Twitter messages have proven useful for early warning in the case of earthquakes and epidemics. So far though such studies have focused on a narrow set of linguistic features without incorporating any knowledge of the underlying network topology. Understanding the dynamics of how the network of users relate to each other, finding hotspots and mapping these to real world locations will form part of a new approach to disaster information systems which our groups plan on collaboratively developing.

Twitter has been active without major disruption during the Fukushima accident. In Tokyo, the surface lines have been down for two days but one mobile network (out of three) has always been in working order, so we really think this scenario is realistic. It is anyway rewarding to monitor social media since it has been proven that the response time in case of disaster is largely lower when compared to public authorities reactions.

Position description

We are looking for a Post-doc whose main objective will be to develop algorithms and knowledge resources that identify communities of need and supply through linguistic and topological analysis as well as grounding of those communities in locations of the real world. Needs will correspond to anxiety about basic resources that naturally arise after disasters such as safety, food, water, medicine, communications and so on. These needs and their corresponding supply chains have been the focus of manual detection in various humanitarian projects such as Ushahidi. The difference we hope to make in this work is to analysis the personal reports automatically using advanced algorithms. In the first instance LaTTiCe and the Collier Lab aim to develop a collaborative project on the topic for two years.

Skills required

- computer science, esp. a strong experience in software development
- machine learning
- natural language processing
- fluent in English
- knowledge or interest in Japanese language
- interest in blogs and new media analysis

https://mail.google.com/mail/u/0/images/cleardot.gif