ICoS-3 Conference Report
From SIGSEM
A conference report by Raffaella Bernardi
Hi, I'm ICoS. Nice to meet you!
I already know something about you. Since you ended up at this web page, I guess you are interested in logic, or linguistics or computation, aren't you? But maybe you don't know anything about me. So I will introduce myself a bit.
I'm a three-year old workshop. I was born in Amsterdam and then moved to a wonderful castle in Dagstuhl. This year I gave my birthday party at Siena. This is a great life for me and my guests: we travel the world to celebrate my birthday! If you want to join us, you know what you have to do. Send your paper to my tutors: it should be on Inference in Computational Semantics ... yes ... that's where my name comes from. Why do logicians and linguists care about inferences? I think it is because we use it everywhere ... that's how I knew something about you!
This year my tutors where Patrick Blackburn and Michael Kohlhase. They did a great job! The guests were from all over the world, and other people were there for birthday parties of workshops older than me. The grand party was called IJCAR because it was an International Joint Conference in Automated Reasoning. Those workshops are good relatives of mine and I think we will get closer and closer, I saw our guests have a lot in common.
I got a bit worried when I saw that the party was going to be in a Hospital and that my guests would have to go through the "Emergency Entrance" to reach me. But they were laughing about it ... though, none of them dared to give their talk using the X-ray standing near the projector!
My party went on for two days from the 18th to the 19th of June. The programme was rich and diverse: besides the standard talks on research topics, we had invited guests (David Israel, Alexander Koller and Ian Pratt-Hartmann), tutorials (given by Peter Baumgartener and Claire Gardent), and demos. Already from the first talk by Ian Pratt-Hartmann you could grasp the high level of the talks, the interest of gathering together results from logic, language and computation, and the close connection between formal and natural language, and between natural language processes performed by computers and human beings. The last invited talk given by David Israel underlined two important aspects which are starting to emerge and which should be pushed forward in the field, namely how to handle real texts and real reasoning tasks. The connection with the "real world" was the Leit motiv of the two days.
A first example of a real problem was given by the junior invited speaker, Alexander Koller. Did you ever hear of computer games, in particular of text adventures? Well, in Saarbrucken computational linguistics students have decided to apply what they are learning from books to improve the performance of those games. Great idea, isn't it? If you want to test it yourself go to this link.
And have you ever had to answer questions asked by a computer? Well ... think of when you have to book a ticket at a travel agency. Aarne Ranta showed how this process can be improved using a proof editor and type theory.
An example of how to deal with real texts was presented by Maarten de Rijke. The sources for the computational analysis where provided by Internet, and the goal was to deal with entailment checking in order to determine the informativity of the given texts. The system presented has been fully implemented and experimentally tested. The method and measures used for the test have been carefully described and a competition has been launched to see which systems get the highest score. Take the ball if you can enter the game!
The internet has been used in other ways as well, for example Sandiway Fong has employed the network of synonym/antonym relations in WordNet to evaluate cases of adjective-verb opposition. Again a connection with the real world has been established and could be further pushed extracting the data from corpora. An essential issue which was present in all the talks is the fact that natural language always involves inference based on the general knowledge of the speakers and hearers. This holds both when playing a computer game (see Alexander's talk) and when parsing sentences involving either presuppositions (see Kristina Striegnitz), or the Frame Problem (see Sandiway who looked at sentences requiring change of colors), and also when answering questions asked either by a human being (see Allan Ramsay) or a computer (see Aarne).
As soon as you are interested in a computational account of such phenomena, three kinds of tools are needed: a parser to analyse linguistic strings, a logic system to model the background knowledge, and an algorithm to get the system running on a computer. Lexicalized Tree Adjoining Grammar, Type Theory, and Lambek Calculus have been shown to be suitable parsers (see Kristina's, Ian's, Raffaella Bernardi's talks); Description Logics, the two-variables fragment of FOL, the five valued logic, Transparent Intentional Logic and McLogic have been used to model the background knowledge (see Alexander's and Kristina's, Ian's, Michael's, Ales Horak's and Jana Sukkarieh's talks); and implementations have been obtained using the Resolution Calculus, information retrieval tools, and other algorithms (see Michael's, Maarten's, Aarna's and Allan's talks).
Besides all these interesting talks each year I get different kinds of presents as well. The guests who have their systems implemented give us a demo. This year the only guest who brought one of these present to me has been Johan Bos. He is a good friend of mine who every year send the invitation to my party around. He delighted us with DORIS, a discourse oriented representation and inference system which translates English texts into Discourse Representation Structures. You can discover which linguistic phenomena it covers going to this link. Please, show me your system if you come to my party next year. I like demos a lot!
Moreover, this year we had something new. As I told you in the beginning, I discovered that my guests shared their interest in computers and reasoning with the big group of people who participated at my relatives' parties --- do you remember ... the automated reasoners?! This was not new to my tutors, though, who had already seen ahead and organized tutorials about automated reasoning for computational semanticists, and computational semantics for automated reasoners. Both tutorials have been given by experts in the area, Peter Baumgartner and Uli Furbach gave the first one, and Claire Gardent, the second. During the former my guests got familiar with tableaux techniques and with the latter my relatives' guests have received a broad and appealing introduction to basic problems and tools of computational semantics, like Higher Order Unification, First Order Model Building and First Order Proving. I heard positive responses to this initiative from my guests, I am curious to know what the others thought of it!
At the end of the two days, another surprise gift has been announced. Elsevier Science made a special offer to my guests for the book which everybody want to have in his bookshelf ..., viz. the Handbook of Logic and Language (LoLa).
Last but not least, I was glad to hear that in honour of Richard Montague --- he is the root of my family's genealogical tree --- one of my guests had the chance to win LoLa for giving the best talk at my party. The winner was Alexander Koller. Congratulations again, Alex!
Finally, a buffet was organized for all of us in the wonderful cloister of the University of Siena. A pity my guests had to stay covered under the porches due to the unexpected rain which started suddenly before leaving the Hospital. Well ... actually I have an explanation for this unbelievable shower we had. If you remember, it rained last year as well, and like it happened this year the thunderstorm started when the name of God appeared in one of the wonderful power point presentation we had. We found this mixture of sacred and profane funny, but apparently not everybody agreed with us. Please, do not try again. I'd like the sun shining during my next parties!!
If you want to read the papers I told you about, you can find them at my homepage. As it happened last year, some of them are going to be published in a special issue of the Journal of Computational Language.
Before saying goodbye to you, I'd like to thank Michael and Patrick for the nice time we had, and the chair of IJCAR, Fabio Masacci, for the excellent organization and the opportunity he gave to me and my guests of getting to know my relatives better.
In closing, I'd like to leave you with this paragraph I read on Maarten de Rijke's slides. I like its message and I hope you share my opinion about it. I hope to see you next time!!
The focus provided by experimental evaluation brings research communities together, forces concensus on what is critical about the field, and leads to the development of common resources, all of which then stimulates further rapid progress. [Kilgariff and Palmer, 2000]
Good bye,
ICoS
>> back to Conference Reports
