Idiap invites applications for several PhD and post-doctoral positions in document-level machine translation and in multilingual speech recognition. These positions are funded by a new 3-year EU Horizon 2020 project in big data, which will develop solutions for monitoring a massive number of media streams through their automatic analysis and aggregation into a dynamic knowledge base.

This announcement is for a PhD position in document-level MT, within Idiap’s Natural Language Processing Group (http://www.idiap.ch/nlp). The successful candidate will be supervised by Dr. Andrei Popescu-Belis and will have the opportunity to collaborate with other members of the NLP Group, who have made significant contributions to document-level MT and information filtering in the recent past.

Description of work

Statistical machine translation systems have shown impressive robustness, adaptability and efficiency. Recently, the community started to consider that document-level knowledge can be used to improve MT at both local and document levels. Numerous open questions remain, regarding the nature of such features and their optimal integration into the training and decoding processes.

The goal of this PhD thesis is to design models for entity-aware machine translation, tailored to the multilingual media streams provided by project partners. These models should improve the translation of entity mentions by hypothesizing story-level coherence and taking advantage of correlations across streams. The models should also accommodate the high input volume characteristic of the project.


Idiap Research Institute


Candidates should have a strong background in computer science, engineering, or a related discipline. Previous experience with statistical MT is an advantage. Candidates should have strong programming skills (e.g. in Python, Perl, Java or C/C++) as well as good knowledge of English as a working language. Knowledge of several other languages is an advantage. A strong academic record and a clear aptitude for creative research will be priority selection criteria.

Specific requirements:

The successful candidate will be registered as a PhD student in the EPFL EDEE doctoral program (typically for 4 years), to which he/she will apply with support from Idiap.

How to apply:

Please mention NLP People as a source when applying

Please fill in and submit your application through the Idiap online recruitment system, by clicking on the position’s title at www.idiap.ch/education-and-jobs.

Tagged as: , , , , ,

About Idiap Research Institute

The Idiap Research Institute is among the most active independent research institutions in information technology.

The Idiap Research Institute, based in Martigny (Valais/Switzerland), is a non-profit foundation that conducts basic research and development in the area of multimedia information management. Idiap's activities can be divided into five complementary research themes and ten application areas.

Idiap was founded in 1991 by the Town of Martigny, the State of Valais, EPFL, the University of Geneva and Swisscom. Since 2008, Idiap has reinforced its connection with Ecole polytechnique fédérale de Lausanne through a joint development plan and a strategic alliance with the "ETH Domain". Several of Idiap's scientists hold an EPFL academic title and are teaching graduate and undergraduate courses. They also supervise students who pursue their doctoral thesis research at Idiap.

Idiap's budget, which amounts to more than 10 million Swiss francs, is 60% financed by research projects awarded following