Post-doctoral position in NLP / text mining

Title: detecting metaphors in scientific discourse
Place: ERIC Lab (Lyon, France)
Supervisors: Sabine Loudcher, Julien Velcin (ERIC Lab), Isabelle Lefort
(EVS Lab).
Duration: 6 months
Salary: about 2000EUR per month (gross salary)

Context
Metaphor is a figure of speech that is broadly used in common language.
It is grounded in the analogical process; it lets a word gets the
meaning generally attributed to another word in a sentence. Metaphor is
very much employed in scientific discourse, in particular in Social
Sciences in which it is at the core of intellectual reasoning. People
think first with those images. This explains why metaphor is a crucial
issue to study, not only in language sciences and cognitive sciences,
but also in geography in which it is employed by many researchers.  The
automatic detection and interpretation of metaphors is a critical task
in Natural Language Processing (NLP) and Information Extraction (IE) as
well.
In the fields of artificial intelligence and NLP, works on metaphor
processing started in the 80s and they give us much information on the
structure and the mechanisms of such phenomena. In the last decade, we
witnessed a technological leap forward with an ever-increasing number of
statistical-oriented techniques. In comparison with more traditional
approaches based on manually-encoded knowledge, these recent techniques
tend to have a greater coverage; they are less precise but gaining in
efficiency and robustness. In this line, several approaches for metaphor
detection have been proposed (Gedigian et al., 2006; Krishnakumaran and
Zhu, 2007; Shutova et al., 2010). They rely on the use of a lexical
resource such as WordNet, TreeBank, etc. Unfortunately, those lexical
resources are mainly available in English, French and Spanish. For the
other languages, either they are of low quality or they don't exist at
all. Yulia Tsvetkov et al. (2013) propose a metaphor detection method
that does not use such a lexical resource.
This proposal lies precisely in the recent trend of using machine
learning for detecting and interpreting metaphors. In this post-doc, we
focus on a specific social science that is Geography.

OBJECTIVE
The post-doctoral researcher will rely on a previous work done in the
ERIC lab and the EVS lab. In particular, a manual encoding has been done
on a small sample of articles and a first implementation based on a
state-of-the-art method based on topic modeling has led to preliminary,
highly promising results. S/he will work on the following tasks:

- complete the implementation of the method of Heintz et al.  (2013) by
  adding some post-processes and testing them on the available corpus
  (code in Python),
- propose new ways to improve the approach, for instance by relying on
  word embedding techniques (several researchers in the ERIC lab are
  currently using those techniques),
- co-supervise the acquisition of a huge corpus composed of articles in
  Geography and the (partial) manual annotation,
- write a collaborative paper to be submitted to an international
  conference
- perform a first temporal analysis of the metaphors (if time allows)


REQUESTED SKILLS
We expect candidates with a PhD in computer science / applied
mathematics, specialized in natural language processing or text mining
with strong programming skills (Python is the preferred language). Some
knowledge in topic modeling would be a plus.

Please send your application including a CV, a cover letter along with
recommendation letters (at least one) to Isabelle Lefort
(isabelle.lefort@univ-lyon2.fr), Sabine Loudcher
(sabine.loudcher@univ-lyon2.fr) and Julien Velcin
(julien.velcin@univ-lyon2.fr).

REFERENCES

Gedigian Matt, John Bryant, Srini Narayanan, and Branimir Ciric.
Catching metaphors. Proceedings of the 3rd Workshop on Scalable Natural
Language Understanding, pages 41--48. 2006.

Heintz Ilana, Ryan Gabbard, Mahesh Srinivasan, David Barner, Donald S
Black, Marjorie Freedman, and Ralph Weischedel. Automatic extraction of
linguistic metaphor with lda topic modeling. In Proceedings of the First
Workshop on Metaphor in NLP, pages 58--66, 2013.

Krishnakumaran Saisuresh and Xiaojin Zhu. Hunting elusive metaphors
using lexical resources. Proceedings of the Workshop on Computational
approaches to Figurative Language, pages 13--20. 2007

Sun Shutova Lin and Anna Korhonen. Metaphor identification using verb
and noun clustering. Proceedings of the 23rd International Conference on
Computational Linguistics, pages 1002--1010. 2010.

Tsvetkov Yulia, Elena Mukomel, Anatole Gershman. Cross-Lingual Metaphor
Detection Using Common Semantic Features. First workshop on Metaphor in
NLP (Meta4NLP). 2013.