The work will explore the design of modules first to analyze the
content of the discussions taking place during a debate, and then
progressively to interact as agents within the on-going discussions.

The analysis part will address issues such as the construction of
semantic maps, accurate visualizations of the ongoing exchanges, and
help for orienting new participants in a debate (rooting). The
analysis will also consider the interactions between the participants
and all kinds of stylistic, semantic, and discursive elements to
identify the key elements of a thread of discussion (to propose
automatic summaries) and to identify the main actors (in terms of
ideas, synthesis, coordination, …)

The interaction part with the participants should progressively
propose richer conversational forms (chatbot), by exploiting the
analysis of the discussion and a good knowledge of the actors
(history, profiles) to get over rigid scripts.

The post-doctoral researcher will have to identify the main issues, to
explore and select the best practical solutions, to complete them with
his/her own developments, and to run some experimentations and
evaluations. Whenever possible and modulo confidentiality issues,
he/she will be encouraged to publish.




– PhD in Computer Science or Computational Linguistics
– strong competences in Natural Language Processing
– competences in Machine Learning (and even better in Deep Learning)
– excellent research track records
– good programming skills

Language requirements:

– good language level in French and English

Specific requirements:

12 months

Educational level:

Ph. D.

Tagged as: , , , , , ,