DCLRS Seminar
The Dublin Computational Linguistic Research Seminars (DCLRS) brings together computational linguistic researchers in Dublin.
Each year, one of the following third-level institutes hosts the seminar series.
- University of Dublin, Trinity College
- University College Dublin
- Dublin City University
- Dublin Institute of Technology
Schedule for Autumn 2008
October 17 Mark Buckley (Edinburgh/Saarbruecken)
October 24 Marilyn Walker (Sheffield)
October 31 Tony Veale (UCD)
November 7 Jonathan Ginzburg (KCL)
November 14 John Kelleher and Brian Mac Namee (DIT)
November 21 Alexander Troussov (IBM, Dublin)
November 28 Rachele De Felice (Oxford)
December 5 Nick Campbell (ATR/TCD)
05 Dec 2008
Venue: Jonathan Swift Lecture Theatre (Arts Building 2041a)
Trinity College Dublin
Time: 16:00, Friday, December 5, 2008
Title:
The Expanding Role of Prosody in Speech Communication Technology
Speakers:
Professor Nick Campbell
Center for Language and Communication Studies
Trinity College Dublin
ATR Spoken Language Communication Research Labs
Acoustics & Speech Processing Department
Abstract:
Speech communication is a uniquely human attribute that plays a
multi-faceted role in human social interaction. At its core from one
point of view lies language and linguistic structure, yet from a more
fundamental point of view we find 'prosody' underlying many levels of
speech communication, serving to signal not just linguistic but also
interpersonal and social information.
Early humans would have had recourse primarily to tone-of-voice for
basic communication but as language use became more sophisticated over
evolutionary time this medium of human interaction became subsidiary
to more sophisticated elements of communcation, though its use did not
disappear entirely.
In the development of technology for processing human speech, the
linguistic element has long been considered prime. This talk will
focus, however, on the 'tone-of-voice' aspects of prosody in social
interaction, tracing their develop[ment in technological research from
a carrier of linguistic information, signalling semantic and syntactic
structure, to that of a social indicator, signalling affective and
interpersonal cues that are equally essential to effective
communication in a social situation.
By thus unravelling the role of prosody in speech, we will trace its
uses from higher to lower levels of sophistication, and suggest some
aspects of prosodic interpretation that might enable a technology for
the processing of interpersonal states and attitudes in addition to
and alongside the processing of propositional content in the speech
signal.
28 Nov 2008
Venue: Jonathan Swift Lecture Theatre (Arts Building 2041a)
Trinity College Dublin
Time: 16:00, Friday, November 28, 2008
Title:
Automatic correction of prepositions and determiners in non-native
English
Speakers:
Dr. Rachele De Felice
Educational Testing Service
Princeton, NJ 08541
Abstract:
In this talk, I present a contextual feature-based approach to the
automatic acquisition of preposition and determiner models of
use. These models can associate prepositions and determiners to their
context with 70.12% and 92.15% accuracy respectively for grammatical
text. I also show how they can be used for the automatic detection of
errors in non-native English writing, and illustrate some of the
challenges presented by this task
21 Nov 2008
Venue: Jonathan Swift Lecture Theatre (Arts Building 2041a)
Trinity College Dublin
Time: 16:00, Friday, November 21, 2008
Title:
Mining of Graph Data in Social Semantic Desktop
Speaker:
Dr. Alexander Troussov
IBM Dublin Center for Advanced Studies
Chief Scientist
Abstract:
I will introduce the Social Semantic Desktop created by the EU project
Nepomuk and will demo IBM components responsible for related item
recommendation. This includes ontology-based term disambiguation,
automatic metadata generation for free texts, recommendation of
related items based on link analysis. I will also provide the
introduction into methods of dynamic local search and ranking for
networked data used in IBM components.
14 Nov 2008
Venue: Jonathan Swift Lecture Theatre (Arts Building 2041a)
Trinity College Dublin
Time: 16:00, Friday, November 14, 2008
Title:
Medical Language Processing for Patient Diagnosis Using Text
Classification and Negation Labelling
Speakers:
Dr. Brian Mac Namee Dr. John D. Kelleher
School of Computing
Dublin Institute of Technology
and
Dr. Sarah Jane Delany
Digital Media Centre
Dublin Institute of Technology
Abstract:
Recently, the i2b2 Center for Center for Biomedical Computing at the
University of Albany organised an natural language processing shared
task that focused on obesity and its co-morbidities. The data for the
task consisted of discharge summaries for patients, a discharge
summary is a textual narrative, semi-structured patient record of all
the care delivered to a patient during their stay. Each of the
discharge summaries was annotated at the document level with
information regarding whether the patient was suffering from obesity
or any of its co-morbidities. The goal of the challenge was to develop
classification systems that could, given a discharge summary, return a
diagnosis of what morbidities the patient was suffering from. This
talk will describes the approach to the challenge adopted by the DIT
AIGroup ( http:www.comp.dit.ie/aigroup ). Based on experimental
results a system was developed which used knowledge-light text
classification using negation labelling and the UMLS medical lexicon.
Funded by
