Orange Labs, the R&D division of Orange, offers 2 post-doc positions
in Natural Language Processing, starting in April 2018.

-Duration: 12 months, starting in April 2018, open until filled.

-Location: Lannion, France

-Contract: fixed term position

-Remuneration: approx. 2,400¤ /month net income (in addition, the
 contract includes health benefits, variable gratification and various
 corporate advantages)

 Applicants will work in the Orange Labs NLP team, composed of ~20
people, which gather various skills, from research on several NLP
suject to the development of actual industrial analytics tools. In the
research domain, besides from feeding the development team with new
algorithms, we collaborate with academic labs, through direct research
contracts or trough collaborative projects (ongoing French ANR funded
projects: PASTEL, DATCHA), and we participate to international
evaluation challenges (in 2017: SemEval on Community Question
Answering and CoNNL shared task on universal dependencies).

 

######################################

# Post-Doc 1: Multilingual Word Embeddings

 

##Topic

Word Embeddings are now considered as the standard level of word
representation in many tasks of Natural Language Processing. Trained
in an unsupervised way, thus being able to take advantage of huge
amount of textual data, they capture semantic and syntactic relations
between words, and can be used in many supervised or unsupervised NLP
tasks. Embeddings estimated on one language are not initially
compatible with embeddings trained on another language. The goal of
the post-doc is to study different solutions to build a unified
representation space for different languages, where semantically
related words from different languages will be close.

Several recent research works have been devoted to bilingual word
embeddings, and methods can be divided into 2 main approaches. One
consists in training bilingual embeddings based on parallel
corpora. The other attempts to map, in a shared space, embeddings
which have been trained independently on different languages. This is
the latter method which will be explored in the post-doc, due to the
unavailability of parallel corpora in the applicative fields studied
by Orange.

 

## Profile:

PhD in computer sciences

Familiarity with language technology

Good knowledge of English and ability to integrate a French speaking
working team.

## Contact Information:

delphine.charlet@orange.com