Albert Gatt
Senior Lecturer
Institute of Linguistics, University of Malta

Bio

I am a tenured Senior Lecturer at the Institute of Linguistics,
University of Malta, a multidisciplinary institute focussing on
theoretical, computational and psycholinguistic research. Prior to
joining the Institute, I did my PhD and a post-doctoral research
fellowship at the University of Aberdeen, where I worked within the
Natural Language Generation group. I have also held a visiting post
within the Tilburg centre for Communication and Cognition (TiCC;
Tilburg University, The Netherlands), working on the project "Bridging
the Gap between Psycholinguistics and Computational Linguistics: The
Case of Referring Expressions".  I take a broad interest in areas
related to language production (by humans) and language generation (by
machines).  Specifically, my research has focussed on the generation
of text from raw data (e.g. in the BabyTalk Project and subsequent
work), as well as referring expression generation (REG). Within the
latter field, my recent interests have been in testing the predictions
of REG algorithms using psycholinguistic experiments, and in
developing stochastic REG models that can account for non-determinism
in human production data. Another of my current interests is in the
generation of narrative from real-world data; specifically I am
interested in questions such as: "How does one discover a story in a
dataset that can serve as a frame for adequate presentation in natural
language?" and "How does one handle temporal data, including temporal
uncertainty, to convey to a reader of a generated text what happened
and when?" A third area of long-standing interest for me is
comparative evaluation, particularly the question of how different
evaluation methods can shed light on different aspects of the quality
of a generated text.

I have co-organised ENLG in Sofia, Bulgaria in 2013, and have served
as co-chair for the Generation and Summarisation track at COLING in
2014. I have also been a member of the NLG Shared Task steering
committee since its inception, and together with Anja Belz, have
co-organised a number of shared tasks, focussing on referring
expression generation, which were among the first such tasks to be
organised in the field.


Joint Statement with Claire Gardent

One recent evolution that is particularly relevant for the NLG
community is the proliferation of readily available symbolic data such
as linked data and large scale open source knowledge bases. There is
also a large body of work on text-to-text processing (including
summarisation, sentence and text simplification, sentence compression)
which raises interesting issues for NLG.  If elected, we will strive
to promote the development of an NLG community working around those
themes e.g., by supporting the organisation of shared tasks (which
could focus on all or some subparts of the NLG process) or by inviting
people from the semantic web and from the text-to-text processing
communities to participate in the SIGGEN conferences (via tutorials,
invited talks, demos, workshops, etc).

To strengthen the ENLG/INLG conferences, we will encourage
participation by a larger audience e.g., by extending these
conferences with workshops and tutorials and by liaising with existing
structures/projects lying at the interface between NLG and other
fields such as for instance, the iV&L European COST action of Vision
and Language and more cognitively oriented fields such as in
particular, computational narratology.

We are also keen to initiate a debate about SIGGEN itself and about
its role in supporting both the INLG/ENLG conferences and the
organisation of shared tasks. While shared tasks are an important
factor in furthering the development of a scientific field, they are
time consuming and require strong individual commitment. We will
endeavour to set up a more permanent infrastructure whose role will be
to encourage and support the organisation of shared tasks for NLG.