IWCS-8 Conference Report
From SIGSEM
A conference report by Caroline Sporleder and Luciana Benotti
Contents |
Fifteen years of IWCS in Tilburg
The Eighth International Conference on Computational Semantics took place from January 7 to 9 (2009) in Tilburg (the Netherlands). As always since it beginnings in 1994, this biannual event was organized by Harry Bunt. And as in previous years researchers from all over the world flocked to the southern Dutch city of Tilburg to hear about the latest developments in computational semantics, engage in lively discussions, plan future collaborations, and enjoy a good share of Dutch hospitality. This year Tilburg greeted its guests with a lot of snow and arctic temperatures, but also a good share of bright sunshine.
Wednesday 7
After a brief opening by Harry Bunt, the conference was kicked-off with an invited talk by Martha Palmer on word senses and sense relatedness. Martha argued that a focus on discrete senses fails to capture the fact that some word senses seem to be more closely related than others, as evidenced by psycholinguistic studies. Hence, clustering related word senses leads to little information loss and may be a beneficial strategy for many applications. The second presentation of the conference was given by Johan Bos, who demonstrated convincingly that superlative constructions give rise to many interesting semantic problems.
After the coffee break, Dan Garette introduced an extension of the NLTK for dealing with semantic processing that he has developed together with Ewan Klein. Allan Ramsay (in joint work with Deborah Field) followed with a presentation in which he suggested that it may sometimes be better to present world knowledge in natural language ("English with variable slots") rather than in some logical formalism. This talk sparked off a lively discussion about the advantages and disadvantages of formal logics for knowledge representation. The morning session was concluded by Wei Chen with a talk about how mental states can be computed from natural language.
After a typical Dutch lunch with sandwiches and milk, the conference continued with a paper by Livio Robaldo and Jurij Di Carlo on disambiguating quantifier scope in Dependency Tree Semantics. Paul Bedaride then presented joint work with his co-author Claire Gardent on semantic normalization of linguistic annotations.
In the final session of the day, Harry Bunt presented his work on semantic annotations and underspecification, arguing that the former should always add information to the source text. As a special typographical treat, this talk featured slides with bullet points in the form of the IWCS logo! After an interesting discussion about what it means for an annotation to "add information to the text", James Pustejovsky came up to the stage to present joint work with Jessica Mszokowics, Olga Batiukova, and Anna Rumshisky on the GLML mark-up language. The formal program of the first day finished with a summary of the ISO-SIGSEM Meeting that had preceded the conference and a short SIGSEM Business Meeting in which Johan Bos encouraged everybody to join SIGSEM (if they weren't already members).
The day was concluded with a trip to the Town Hall, which was built as a royal palace by King William II of the Netherlands, who, alas, died before the building was completed. During the guided tour of the building we visited numerous wedding rooms as well as a room in which the young Vincent van Gogh enjoyed early painting lessons. Our competent tour guide also provided us with many amusing facts and anecdotes about Tilburg's royal past. After the tour we settled down to a well-deserved drink during the reception.
Thursday 8
Thursday's invited speaker was Ido Dagan, who argued that the computational semantics community should follow the example given by the communities working on morphology and syntax. It's time for the semanticists to have a unified and comprehensive engine which addresses the semantic tasks required by different applications. Ido Dagan discussed a first proposal of the engine API, and suggested Recognizing Textual Entailment (RTE) as its basic semantic task.
After the invited talk, Bill MacCartney presented joint work with Chris Manning on extensions to natural logic that yield performance gains in the RTE challenge.
The second session started with a talk by Fabio Celli, who presented joint work with Malvina Nissim on the automatic identification of Italian complex nominals. Then followed a talk by Georgiana Dinu, who, together with her colleague Rui Wang, looked at the role of inference rules for RTE. The morning ended with a lively session of several flash presentations that were meant as teasers for the following poster session which was held during the lunch break and lead to many interesting discussions.
The afternoon session featured two talks. The first was by Luciana Benotti who talked about the treatment of comparative constructions in dialogues (joint work with David Traum). In the final talk of the day Katrin Erk presented her work on semantic vector space models and on modeling hyponymic inferences in such a space.
After the last talk of the day, it was time for the conference excursion. This year a bus took all of us to the neighboring city of Breda. After a short trip across a completely white landscape covered by soft powdery snow, we arrived in a no less white Breda. Even the most assiduous participants of IWCS said they didn't remember such a white and cold (but sunny!) IWCS ("we should have brought the scarf from the last IWCS" was a thought that crossed many minds). We warmed up a little in the tourist office where we met the guides that accompanied us around the city during the following couple of hours.
The visit included the beautiful municipal park, the main church of the city which is an impressive Renaissance building, the Breda Béguinage that has a long religious history and still nowadays serves as residence for single women, and the castle of Breda that currently houses the royal military academy. By the time we arrived in the castle the cadets were having dinner and the sun had already set, the tour was over and it was time for some warming up drinks. The animated discussion that followed in a bar was only interrupted by an enthusiastically received "time for dinner!!". Dinner that we all enjoyed à la carte at a cozy restaurant.
Friday 9
Massimo Poesio, the invited speaker of the last day, talked to us about an ambitious project that aims to create a 1-million-word annotated corpora using web volunteers. As you can imagine many volunteers are needed, if you want to be one of them try Phrase Detectives, a collaborative game that collects judgments about anaphoric annotations. The resulting resources are being stored in The Anaphoric Bank, a club created to facilitate resource sharing among researchers working on anaphora. As "unrelated news", Massimo Poesio also announced a new international journal on Dialogue and Discourse.
Patrick Saint-Dizier presented joint work with Isabelle Dautriche about using action theories to formalize procedural texts (such as recipes) in order to integrate them and improve the success rate of the procedure goal. Before the coffee break, Stephen Pulman thanked Harry Bunt for eight successful editions of IWCS. This was one of many times during the workshop that the topic came up; a feeling of closing an era and beginning a new one was in the air.
The coffee break that followed, served to continue discussions that have started during the question time, while enjoying chocolates specially designed for IWCS! Recomforted and warmed up (on a Friday that was even colder than the Thursday!), we listened to Anna Lobanova and Jennifer Spenader talk on their empirical findings of certain discourse markers, which support the distinction of the different types of contrast relations recognized in standard Rhetorical Structure Theory. Olga Petukhova, reporting joint work with Harry Bunt, continued talking about discourse markers concentrating on the multiple communicative functions that single discourse markers have in dialogue. It was motivating to see that (maybe inspired by the resource sharing atmosphere started by Massimo Poesio in the early morning) there were many constructive comments and offers on the resources used for these two empirical works.
The last long presentation of the morning was given by Ruth Kempson (whose co-authors are Eleni Gregoromichelaki, Andrew Gargett, Yo Sato and Christine Howes) who presented Dynamic Syntax as a suitable framework for modeling phenomena that may seem to be uniquely constitutive of dialogue (such as clarifications) but are in fact based in mechanisms of apposition equivalently usable in monologue; maybe dialogue doesn't need special grammars after all?
This very intensive conference ended with many 3-minutes presentations and posters (and of course, lunch) on very diverse topics that made evident how much the area of computational semantics has changed and grown in the last years.
IWCS says "goodbye" to Tilburg
Fifteen years after its birth, IWCS says a nostalgic "goodbye" to Tilburg and to all the people who have made IWCS possible over the years. It's time for IWCS to travel around the world, now organized by Johan Bos and Stephen Pulman, and to continue to be a fruitful opportunity to meet the enthusiasts in computational semantics.
>> back to Conference Reports
