IWCS-7 Conference Report
From SIGSEM
A conference report by Jennifer Spenader
Contents |
This year's workshop illustrated rich tapestry of field
From the 10th to the 12th of January (2007), the International Workshop on Computational Semantics held it's seventh meeting in Tilburg, the Netherlands. The conference is the brainchild of Harry Bunt, who started the series to encourage work in this budding field.
The meeting was held on the ground floor of the Dante building in the heart of the Tilburg University campus. As participants arrived and registered they were presented with an impressive proceedings, 379 pages long (!), as well as a stylish scarf, woven from the pattern of an 18th century garden, and also the symbol of the workshop. This scarf was made for the workshop by the textile museum in Tilburg, which was also the excursion destination for the workshop, but more on that later.
Wednesday, January 10th, 2007
The first talk was given by invited speaker Johan Bos who is no stranger to this series of workshops. Johan gave an introduction to the Recognizing Textual Entailment (RTE) shared task that he had participated in, and strongly encouraged other members of the IWCS community to get involved. Many of the systems still don't know how long a piece of string is, achieving slightly better than 50% using knowledge poor techniques. If more computational semanticists would get involved perhaps we could see faster advances.
Johan's talk was followed by Rafaella Bernardi, reporting on joint work with Diego Calvanese and Camilo Thorne on the possibility of using a controlled English (DL-Lite) to access databases with a fragment of controlled English that has felicitous computational properties. Rafaella's talk was followed by a coffee break.
After the coffee break, where participants inventoried the different color coded cookies available, Ariel Cohen gave a talk outlining how to formalize the linguistic constraint Dont Overlook Anaphoric Possibilities (DOAP; first in Williams, 1997) in Default Logic. Thereafter Dan Modovan presented a paper on joint work with Marta Tatu on automatically recognize intentions in text, with results of experiments they did on using support vector machines to learn models of intentions in real texts. This was followed by a presentation of work by Dennis Spohr, Aljoscha Burchardt, Sebastian Pado, Anette Frank and Ulrich Heid on annotating a corpus for lexical information in a Description Logic-based model that offers a number of advantages beyond XML-based representations, in particular the ability to be able to check consistency and abstract over the corpus annotations. This representation also offer more possibilities for querying the corpus.
The morning session ended and we were all treated to lunch, in the form of quite fancy Dutch "broodjes" (sandwiches), served with the characteristically Dutch drinks of milk and buttermilk! After recharging everyone was ready for the afternoon session, kicked off with a talk by Harry Bunt himself, who presented joint work with Amanda Schiffren on standardization of dialogue act coding, something greatly needed. This talk was followed by at talk given by Olga Petukhova, on joint work also done together with Harry Bunt, on the advantages of multidimensional approaches to dialogue act annotation, where a single utterances can be considered to have more than one function in the dialogue. This was followed by a talk by Edward Loper on joint work with Szu-ting Yi and Martha Palmer on mapping between PropBank and VerbNet, two lexical and conceptual resources. Since the many resources that exist have quite different representation formats, mapping endeavors like this are extremely useful to make using information from more than one resource more practical. This was followed by a tea break.
The session after the tea break included two talks about time. Takashi Koga and Satoshi Tojo presented work on Tense and Aspect in Polymodal Interval Temporal Logic. After this Ian Pratt-Hartman gave a timely talk about a subset of TimeML with more advantageous properties that would make it more suitable for marking up narrative texts. Unfortunately because our presence was required at the Town Hall reception at 16:45 and because we were to be transported in the luxury of our own bus, the temporal logic speakers were not allowed to deviate temporally from the schedule, so several potential questions had to be relegated to the break. Following both talks Harry Bunt informed everyone about the work that the SIGSEM working group was engaged in before the workshop, which included important topics such as the standardization of discourse acts. Harry also brought up a few points of SIGSEM business, in particular about a number of vacancies on the board. We then adjourned to quickly make our way to our waiting bus.
And where did the bus take us? Well it seems that the IWCS-7 organizers were able to pull strings and get us invited to a reception at the town hall in downtown Tilburg. Unfortunately the weather didn't cooperate, and at the point when we needed to leave the bus to make our way across the street to the town hall entrance, we were faced with a torrential downpour. New alliances formed with those with foresight enough to bring an umbrella, and en masse we braved our way to the Town Hall, only to quickly discover that the entrance we chose was not the correct one! After Harry left to see where we were meant to enter, we were lucky enough to find a staff member with keys that could lead us to the reception area. What originally was an error actually turned into an advantage, as we were able to see some of the beautiful interiors of the former palace that is part of the current Town Hall.
In the reception room, a town alderman gave a quite modest speech about how much he hoped we would like Tilburg. We were then treated to refreshments and "hapjes", tiny sandwiches and cracker treats, and I think every participant enjoyed a refreshment and a hapje. As the noise level climbed it was clear that everyone was having an enjoyable time.
Thursday, January 11th, 2007
The following morning started with an invited talk by Paul Buitelaar on the Semantic Web for natural language Paull's talk discussed in particularly the use of multingual ontologies in the semantic web endeavor. This was followed by a talk by Staffan Larsson, on how meaning changes could be accounted for in linguistic theory, but also with thoughts on how the creative use of non-standard or original meanings or senses for existing linguistic terms in dialogues are understood by dialogue participants in real time, suggesting how this might be dealt with in a dialogue system. The session was then followed by a coffee break where many participants began deep discussions of ideas related to computational semantics.
The coffee break was followed by two papers with solo authors: Sylvai Pogodalla presented work on scope ambiguity using a direct approach within an extension of an abstract categorical grammar (ACG). cope ambiguity was also the topic of Alistair Willis' paper on how to extend hole semantics to deal with the interpretation of quantifiers interacting with coordinated NPs. The session was stitched up with a large number of short papers and project abstracts. The flash presentations were sufficient to whet the appetites of numerous participants so that during the lunch that followed, the associated posters were well visited.
The informative lunch was followed by an afternoon session kicked off with a presentation by Claire Gardent, who presented an informative comparison of what happens when a lexicalized reee Adjoining Grammar (LTAG), Feature-based TAG meets each of three different semantic formalisms: Glue Semantics, Flat semantics and lambda-semantics. How LTAGs can be used with a semantic formalism is the subject of the second paper in the session, Phillipp Cimiano, Annette Frank and Uwe Reyle's paper on how to incorporate Underspecified Discourse Representation Theory (UDRT) with LTAGs, given the special way in which LTAGs treat adjunction. Discourse Representation Theory was then the topic of the third speaker, Rui Chaves who presented a method to deal with pluralities in DRT. Because model checking with plurals easily becomes unmanageable, creative ways to deal with model checking need to be developed.
After this session we all hurried to board the bus for the conference excursion. This years excursion was to the Tilburg Textile Museum. As Tilburg used to a be a center of textile production, this was an appropriate destination. Divided into three groups, we were guided through the museum and were able to see examples of textile processing machines from years gone by as well as modern machines that can create material from patterns entered by the user in computer. Participants in the excursion were also able to see the weaving machine that produced the workshop scarf. It was important to stay close to our guides because above the noise of the weaving machines it was often hard to hear and no one wanted to lose the thread. It is clear that the field of textile production has developed rapidly with the introduction of computerized looms, knitting machines and laser cutters, just as the field of semantics has rapidly developed with the introduction of computational resources and application! Such a striking resemblance!
The conference dinner was held on the top floor of Humphreys restaurant in downtown Tilburg. We began with drinks and the gradually increasing noise level was clearly indicative of the enjoyment of the participants.
Friday, January 12th, 2007
Friday began with a talk by Glynn Morill on joint work with Mario Fadda and Oriol Valentin on Nondeterministic Discontinuous Lambek Calculus, presenting a solution to a long pondered problem. In more than one way you could say that Morill, Fadda and Valentin have now wrapped up the problem of wrapping, which has numerous linguistic applications. This work was followed by a talk by Stephen Pulmann on why so many results of formal semantic analyses have yet to be implemented in computational semantics, illustrating the difficulty with adjectival comparison. After his talk it was clear that computational semantics is more difficult than formal semanticists expect it to be. Pulman's talk was followed by Jennifer Spenader and Gert Stulp's paper on the feasibility of using antonym relations to identify the source of contrast in both contrastively marked and non-contrastively marked sentences.
After a short coffee break, where many cookies were enjoyed, were returned for the final session. The first talk after the break was by Derrik Higgens, on joint work with Jill Burstein, both from Educational Testing Service. Derrik presented their experiments on automatically determining coherence in essays. This work in particular illustrates the rich tapestry that computational semantics is becoming. The talk was followed by the final invited talk of the conference, by Makoto Kanazawa on a method of grammar induction using lambda calculus. Grammar induction is a fascinating topic, with many working at ways to automatically derive syntactic structures. Because Kanazawa is a semanticist/logician, we could say his grammar induction coat is cut according to his cloth.
Finally the session ended with a number of flash presentations of short papers and project abstracts, with many interesting new ideas. Finally, Harry Bunt thanked all the participants and the organizing crew for their hard work. We then adjourned for lunch, where besides the large pile of tempting broodjes the many intriguing posters from the flash presentation session were waiting to be consumed.
Group Picture
Front row, from left to right: Jennifer Spenader, Katrin Erk, Philipp Cimiano, Paul Buitelaar, Ariel Cohen, ???, Ian Pratt-Hartman. Second row, from left to right: Michael McCord, ??, ??, ??, ??, ??, ??, Harry Bunt, Kiyong Lee, Raffaella Bernardi, Anette Frank, Glynn Morill, ??, ??, ??, Dan Moldovan, Oriol Valentin. Third row, from left to right: Fred Landragin, Makoto Kanazawa, ??, Manual Alcantara, Gerald Penn , ??, Steve Pulman, Koiti Hasida, Claire Gardent, ??, Thorsten Trippel, ??, Rui Chaves, Livio Robaldo, Alistair Willis, Staffan Larsson, ??, ??. Back row, from left to right: ??, Ralf Klabunde, ??, ??, Johan Bos, ??, Reinhard Muskens, ??, ??, Crit Cremers, Raquel Fernandez, ??, Sylvain Pogodalla, Rainer Oswald.
Nothing to push under the carpet
The IWCS-7 was a great success, and if you want to become part of the rich fabric that the field of computational semantics is becoming, keep in mind that the next event will probably be in 2009, so start preparing your papers now!
>> back to Conference Reports
