261 days ago on emploi.epfl.ch

Ph.D position on "Deep learning methods for understanding texts"

EPFL - Ecole Polytechnique Fédérale de Lausanne

  • Work region
  • Sector
  • Employment type
  • Position

 Please refer to JobSuchmaschine in your application

Ph.D position on "Deep learning methods for understanding texts"

Ph.D position on “Deep learning methods for understanding texts”

Ph.D position on “Deep learning methods for understanding texts”

As part of a FNS project focusing on the effects of algorithm of natural language, we open a Ph.D position on the subject of “Deep learning methods for understanding texts”.With the advancement of so-called deep learning techniques based on temporal convolutional networks, new approaches to text understanding are currently showing increasing performances. The goal of this research is to explore how these techniques can be used to study an hypothetical rapid evolution of language characterised by the systematic mediation of algorithms in text production.

Requirements:Master degree in computer science. Prior experience with deep learning methods is a plus. Openness to interdisciplinary work with humanists, as well as fluent spoken and written English.

To apply for the position:please email your scientific resume to Prof. Frederic Kaplan, Email: frederic.kaplan( at)epfl.ch. Student candidates for PhD must be admitted to one of the doctoral school at the EPFL (EDIC, EDAR, EDMT) to which they need to apply directly. http://phd.epfl.ch/

The salary will be according to the EPFL standards.

The starting date is: February 2016.

About the lab:The Digital Humanities Laboratory (DHLAB), founded in 2012 by professor Frédéric Kaplan develops new computational approaches for rediscovering the past and anticipating the future. Projects conducted at the lab range from reconstruction ancient cities to studying how algorithms transforms the way we write. TheVenice Time Machine, DHLAB's flaghship project conducted in partnership with the Ca'Foscari University ambitions to digitize 1000 years of historic records and make them searchable.

(Edited 19.01.2016/vb)