European Masters Program in
Language and Communication Technologies
Language and Communication Technologies Colloquia
The Colloquia will take place every other Thursday during the
months of December 2004-June 2005 accordingly to the calendar
below.
All seminars are held in English, unless noted differently.
For more information, please contact Raffaella Bernardi or Paolo Dongilli
| Date | Speaker | Affiliation | Title | Abstract |
|---|
| December, 16th |
Paolo Dongilli |
KRDB, FUB |
Rule-based part-of-speech tagging |
Abstract
|
| Winter Break |
| January, 13rd |
Judith Knapp |
EURAC |
The application of computational linguistics tools for computer assisted language learning: Experiences with WordManager |
Abstract
|
| January,27th |
Marcello Federico |
ITC-IRST |
The Statistical Approach to Machine Translation |
Abstract
|
| Feburary, 10th |
Gianni Lazzari |
ITC-IRST |
Spoken Language technologies: from research to market |
Abstract
|
| February, 24th |
Emanuele Pianta |
ITC-IRST |
Natural Language Generation |
Abstract
|
| March, 10th |
Diego Giuliani and Fabio Brugnara |
ITC-IRST |
Automatic Speech Recognition |
Abstract
|
| March, 24th |
Bernardo Magnini |
ITC-IRST |
Web as a Corpus |
Abstract
|
| April, 7th |
Fabio Pianesi |
ITC-IRST |
User-centred design of an affective computing interaction paradigm for an adaptive multimedia mobile guide for museums |
Abstract
|
| April, 21st |
Luca Dini |
CELI s.r.l. |
Information Extraction: Market Demand and Technological Syncretism |
Abstract
|
| May, 26th |
Alessandro Lenci |
Università di Pisa |
Lexical Resources: design and acquisition |
Abstract
|
| June, 9th |
Guenther Neumann |
LT-Lab DFKI |
Cross-lingual Open-domain Question-Answering from Unstructured
Texts |
Abstract
|
| June, 16th |
Fabio Pianesi |
ITC-IRST |
Ethnography of small groups for the design of a co-located multimodal system to support meetings |
Abstract
|
16th December, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Rule-based part-of-speech tagging
Paolo Dongilli, KRDB, FUB
|
|
Part-of-speech tagging is the area of Computational
Linguistics that deals with the automatic assignment of appropriate
grammatical descriptors to words in a text. There are various
approaches to tagging and usually statistical techniques have been
more successful than rule-based methods. The aim of this seminar is to
present a simple rule-based part-of-speech tagger showing how it
automatically acquires its rules and tags with accuracy comparable to
stochastic taggers.
|
13rd January, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
The application of computational linguistics tools for computer assisted language learning: Experiences with WordManager
Judith Knapp, EURAC
|
|
|
|
Computer-assisted language learning (CALL) is a research
field which
explores the use of computational methods and techniques for language
learning and teaching. Recently, an increasing number of language
learning systems have been developed which adopt Computer Linguistics
(CL) tools. I will start with a short review about how these tools are
used and to what extend it has been successful. Then I will present
the ELDIT program. The ELDIT project started at the European Academy
of Bolzano in 1999. ELDIT is an electronic learners' dictionary for
the Italian and German languages and includes also the texts for the
exams in bilingualism. Adding interactive exercises and a tandem
learning feature is planned for the future. ELDIT includes CL tools
to provide educational content in an innovative and meaningful way. In
particular I will speak about WordManager, a reusable morphological
database developed at the University of Basel. Due to WordManager we
can for instance provide the entire inflection paradigm (declination
or conjugation) for each dictionary entry. Realizing such features
requires a close co-operation between linguists, computational
linguists, language teachers and computer scientists. I will conclude
our talk by outlining some experiences about this interdisciplinary
collaboration.
|
27th January, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
The Statistical Approach to Machine Translation
Marcello Federico, ITC-IRST
|
|
|
Machine Translation (MT) is one of the oldest and most ambitious
challenges taken on by computer science. In this lecture, I will
introduces the problem of MT and review the history and main
approaches developed during the last 50 years. Then, I will focus
attention on the currently most successful approach, which is based on
statistical models developed in the early 90's at IBM. Statistical MT
exploits so called word-alignment models which can be automatically
trained from parallel corpora, i.e texts provided with human
translations. Finally, I will overview performance evaluation methods
for MT and report results of recent international evaluation campaigns.
|
10th February, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Spoken Language technologies: from research to market
Gianni Lazzari, ITC-IRST
|
|
|
Spoken language technologies
are already used in very useful applications as automatic telephon
services, dictation machines and data entry for medical and
professional reporting New technology is almost ready for affording
more challenging applications like broadcast news trascriptions, and
spoken document access.
In the talk some of the forementioned applications will be
presented. Moreover the typical problems encountered and the methods
used during the development of the application together with the key
roles played in the business chain, will be discussed.
|
24th February, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Natural Language Generation
Emanuele Pianta, ITC-IRST
|
|
|
After a short overview on the main issues at stake in the field of
automatic Natural Language Generation (NLG), we will introduce a more
specific topic, that is how to enforce robustness in NLG systems.
Robustness is usually not considered as an issue for NLG. This is
explained by the fact the, unlike what happens with natural language
analysis systems, e.g parsers, the input to NLG systems is supposed to
be some formal and well-formed representation, such as a semantic
representation or the content of a data-base or knowledge-base. We
will show that this assumption may be wrong for a generation component
embedded in an interlingua-based machine translation system. We will
illustrate how the robustness issue is tackled by XIG, an
Interlingua-to-Italian generation component used in the NESPOLE!
machine translation project
|
10th March, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Automatic Speech Recognition
Diego Giuliani, Fabio Brugnara, ITC-IRST
|
|
|
The seminar will give an
overview of the different aspects of automatic speech recognition,
from signal processing to probabilistic modeling and decoding. After
posing the problem, motivating the statistical approach, the overall
architecture of a speech recognition system will be presented. For
each components, the basic elements will be described, so as to
clarify its role in a complete speech recognition system.
Most popular methods adopted for modeling and decoding the speech
signal will be briefly introduced together with some elements of
phonetics. In particular, some principles of acoustic modeling, based
on Hidden Markov Models, and statistical language modeling will be
presented.
|
24th March, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Web as a Corpus
Bernardo Magnini, ITC-IRST
|
|
|
|
The Web is an immense, multilingual, freely available
corpus. As with
other large new corpora, computational linguists have been stimulated
by its presence.
Recently there has been a growing interest in techniques that take
advantage of Web redundancy for automatically extracting linguistic
information in order to model language variability. As an example,
approaches have been proposed to collect, on a large scale, topic
signatures, examples for word senses, entailment relations and
paraphrases. In addition, the Web has been exploited as a huge
repository of world knowledge: named entities, semantic relations
among entities, facts about a certain topic are first mined and then
organized and made available for further processing.
The seminar will overview of the main approaches that use the Web as a
corpus for different aims in computational linguistics. Both positive
experiences and open problems will be addressed.
|
7th April, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
User-centred design of an affective computing interaction paradigm for an adaptive multimedia mobile guide for museums
Fabio Pianesi
|
|
|
|
In this seminar, we will present a case-study of the user-centered design of a personalized mobile guide. The design involved a team which comprises graphical designers, engineers and psychologists.
A first attitudinal study was performed to assess the impact of some adaptivity dimensions for a mobile guide in a museum setting, namely the location awareness, the degree of control over follow-ups, the content adaptation with respect to user interests and the content adaptation with respect to history of interaction. The study involved forty subjects exposed to simulations of different situations in a museum setting and then asked to score the simulated systems according to a number of dimensions. The scores were then correlated with some personality's traits of the subjects with the aim of assessing which one of these traits influence the acceptance or the rejection of the different dimensions of adaptivity.
Then, a first prototype of a personalized guide was designed and implemented. The first stage of user evaluation revealed the particular implementation of the degree of control over follow-ups dimension was misleading. A small user study was therefore run in order to inform the subsequent re-design phase. A technique called Action-retrospection protocol consisting in an interview with the subject while using the system was used.
The re-design of the interface brought to a completely new paradigm based on the broad notion of affective interaction. We posed that an interaction based on expressing affective attitude toward the service provided by the system may improve usability of an interface in particular when, like in museums, the technology should not hinder the "real" experience.
This paradigm required the design and testing of a new widget, called the like-o-meter During the re-design phase a number of other small studies using these technique were performed.
Finally, a new prototype was implemented a larger user study has now been started.
|
21st April, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Information Extraction: Market Demand and Technological Syncretism
Luca Dini, CELI s.r.l.
|
|
|
The talk will describe the basic goals of the wide range of
technologies that are nowadays classified under the label of
"Information Extraction". It will become evident that what
characterise this technology is not a set of algorithms or techniques,
but rather the market demand for the transition from unstructured to
structured information, i.e. from textual documents to some relational
representation of certain concepts. Under this view, we will show how
in real-life applications the traditional finite state automata and
finite state transducers framework is usually enriched by "external"
technologies ranging from deep grammar processing to statistical
methods.
|
26th May, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Lexical Resources: design and acquisition
Alessandro Lenci, CNR, Pisa
|
|
|
|
In language technologies, the task of providing the basic semantic description of words is entrusted to lexical resources, which aim at making word content machine-understandable. Since meanings live and arise in linguistic contexts, it is necessary to take into account how semantic information emerges from the actual textual data, and how the latter contribute to meaning formation and change. Consistently, computational lexicons can not only be conceived as repositories of semantic descriptions, but rather at most as core set of meanings that need to be customized, tuned and adapted to different domains, applications and texts. The talk will discuss the impact of these issues on the design and development of lexical resources. A special emphasis will be put on the connections between computational lexicons, terminology repositories and ontologies, as well on automatic methods to acquire lexical information from texts"
|
9th June, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Cross-lingual Open-domain Question-Answering from Unstructured
Texts
Guenther Neumann, LT-Lab DFKI
|
|
|
|
Domain-open Question-Answering (ODQA)-systems is a very active
research field in which methods and algorithms from different areas like
Information Retrieval (IR), Information Extraction (IE) and Natural
Language Processing
(NLP) are combined in novel ways. ODQA-systems receive NL-questions as
input (and
not keywords), analyse huge sets of free texts and return exact answers
as output (and
not documents). Recently, research in cross-lingual ODQA is emerging as
a new topic, i.e.,
the development of ODQA-systems that receive NL-queries in one language
(e.g., German)
and extract exact answers in documents of another language (e.g.,
English). In this talk
I will give a brief overview of the major scientific questions and
challenges
of cross-lingual ODQA. I will then describe the methods and
technologies that have
been developed at the LT-lab of the DFKI, viz. robust NL-query analysis
in open-domains,
hybrid methods for the translation of NL-queries and query expansion,
and strategies for the multi-dimensional annotation of large sources of
unstructured texts.
|
16th June, 16:00-17:00 - Faculty of Computer Science, FUB, Seminar Room (first floor left)
Ethnography of small groups for the design of a co-located multimodal system to support meetings
Fabio Pianesi, ITC-IRST
|
|
|
|
In this seminar, we will present a case-study of the user-centered
design of a personalized mobile guide. The design involved a team which
comprises graphical designers, engineers and psychologists.
A first attitudinal study was performed to assess the impact of some
adaptivity dimensions for a mobile guide in a museum setting, namely
the location awareness, the degree of control over follow-ups, the
content adaptation with respect to user interests and the content
adaptation with respect to history of interaction. The study involved
forty subjects exposed to simulations of different situations in a
museum setting and then asked to score the simulated systems according
to a number of dimensions. The scores were then correlated with some
personality's traits of the subjects with the aim of assessing which
one of these traits influence the acceptance or the rejection of the
different dimensions of adaptivity.
Then, a first prototype of a personalized guide was designed and
implemented. The first stage of user evaluation revealed the particular
implementation of the degree of control over follow-ups dimension was
misleading. A small user study was therefore run in order to inform the
subsequent re-design phase. A technique called Action-retrospection
protocol consisting in an interview with the subject while using the
system was used.
The re-design of the interface brought to a completely new paradigm
based on the broad notion of affective interaction. We posed that an
interaction based on expressing affective attitude toward the service
provided by the system may improve usability of an interface in
particular when, like in museums, the technology should not hinder the
"real" experience.
This paradigm required the design and testing of a new widget, called
the like-o-meter During the re-design phase a number of other small
studies using these technique were performed.
Finally, a new prototype was implemented a larger user study has now
been started.
|
