3-year PhD position in Computational Models of Analogy Making in Natural 
Language (IRIT and University of Toulouse, France)

We invite applications for a fully funded PhD position for 3 years at the IRIT 
laboratory and the University of Toulouse, Paul Sabatier, France, in the 
context of the recently funded project AT2TA on Analogy Making. 

Analogy Making is a remarkable cognitive capability during which similarities 
and differences between two parallel situations are exploited in order to draw 
a common "essence" allowing us thus to categorize an object or a particular 
situation to a preexisting concept or create a new one. Reasoning by analogy  
allows us thus to transfer our understanding of a previous situation to a new 
one and appropriately adapt it. It has been argued that analogy making lies at 
the core of cognition (Hofstadter 2001) and has recently drawn the attention of 
Deep Learning pioneers (Chollet 2017, LeCun 2022). 

In Natural Language Processing analogies usually take the form of a quadruplet 
a:b :: c:d traditionally expressed as "a is to b as c is to d" (for example, 
Paris is to France as Berlin is to Germany). Most of the extant work considers 
a, b, c, d as word embeddings and then relies on geometrical properties of 
those embedding in a higher dimensional space in order to recognise a 
quadruplet as an analogy or to generate a d such that a:b :: c:d forms analogy 
given a, b and c. Despite the importance of analogies, most works in NLP do not 
consider analogies between sentences and do not concentrate on the underlying 
latent relations that form the common essence between pairs (a,b) and (c,d). 
The successful PhD candidate will work on computational models which can 
identify analogies between sentences or even bigger chunks of text with a 
particular focus on the identification of common latent relations which are 
also an essential part for an explainable AI. 

The successful candidate should hold a Master's degree in computational 
linguistics or computer science or cognitive science and has prior experience 
in word embedding models or deep learning approaches in general. The candidate 
should have strong programming skills and expertise in machine learning. The 
position is affiliated with the IRIT laboratory at Toulouse and there will be 
frequent interactions with researchers at the Loria laboratory in Nancy in the 
context of the AT2TA project. 

Applications will be considered until the position is filled, but applicants 
are encouraged to apply as early as possible since applications will be 
considered at the moment of reception. Applications, in English or French, 
should include a detailed CV, a letter of motivation and at least two 
recommendation letters. Applications should be sent to Stergos Afantenos 
(stergos.afantenos at irit.fr). The starting date of the position will be in 
January 2023, but can be filled earlier if need be.

More information can be found here: 
https://cloud.irit.fr/index.php/s/gegXukURCTNL2ks

_______________________________________________
uai mailing list
uai@engr.orst.edu
https://it.engineering.oregonstate.edu/mailman/listinfo/uai

Reply via email to