We are looking for a master's intern at CEA Saclay (Paris area).
Please find below the information of the internship.

Location: CEA, Université Paris-Saclay, Palaiseau, France
Duration: 5-6 months, starting from April 2025 (flexible)
Application deadline: January 31st, 2025
Supervisors: Gaël de Chalendar, Nasredine Semmar, Galo Castillo


Context:
Turn-taking management in the context of dialogue systems studies how
agents manage their participation in dialogues. Modeling turn-taking
consists in detecting when an agent should intervene to contribute in a
conversation. Turn-taking modeling in multi-party conversations (MPCs),
where more than two interlocutors participate in the dialogue, has
received little attention in comparison to dyadic scenarios, where only
two interlocutors communicate, e.g. the user and an agent. The lack of
works in MPCs is partly due to the limited amount of annotated corpora
available to study the subject. In addition, most prior work in MPCs
use corpora without explicit turn-taking-related annotations. Such
studies mostly rely on end of turns, which is not suitable for training
systems that do not actively participate in conversations, e.g.
dialogues with a high amount of out-of-scope utterances.

Objectives:
The objective of this internship is to develop an annotated corpus of
spontaneous (i.e. non-scripted) multi-party conversations, with a focus
on turn-taking-related annotations. Annotations are expected to be used
to build virtual assistant systems. Annotation types may include but
are not limited to turn-taking, out-of-scope intent detection,
goal-tracking, etc.

Work to be carried out will include a literature review on methods for
multi-party corpora creation; review and analysis of existing corpora
for turn-taking modeling in multi-party conversations; creation of a
corpus of spontaneous multi-party dialogues for turn-taking modeling
and other dialogue system sub-tasks. The intern is expected to provide
a scientific approach to the development of the corpus.


Qualifications:
-   Pursuing a master's degree in Linguistics or similar
-   Professional proficiency in English
-   Basic knowledge of Python to analyze text data
-   Experience with data annotation tools is a plus but not required


How to apply:
To apply to this internship, please send an email with your CV and a
cover letter to Gaël de Chalendar (gael.de-chalendar@cea.fr) and
Nasredine Semmar (nasredine.semmar@cea.fr).