Constraint Solving and Language Processing: First International...

Features

• Cover Type: Paperback with 205 pages
• Published by: Springer
• Edition: 1st Edition July 21, 2005
• Written in: English
• ISBN 10 Number: 3540261656
• ISBN 13 Number: 978-3540261650
• Book Dimensions: 8.9 x 6.1 x 0.3 inches
• Weighs: 11.2 ounces

Reader Reviews
I can still remember the excitement I felt back in 1990--my major professor had handed me the paper "Constraint Dependency Grammar" by H. Maruyama, and after reading it is seemed like a whole new world was open. Hard problems seemed to suddently become tractable--it was obvious to us tremendous power and flexibility of representing the process of parsing a sentence as that of constraint satisfaction. Problem was, the rest of the world viewed the work with a big, collective *yawn*. Dependency grammars were SO passe'--constituency, not dependency, was orthodoxy! Who needs new parsers? Isn't that just re-inventing the wheel? After all, context-free grammars were well-understood and already had very efficient parsers, so why do we need an entirely new grammatical formalism? About this time it started to be fashionable to induce grammars from corpora. Much effort and money was spent to find ways to learn grammars. Even here, Dependency grammars seemed to have an edge--back in '95 I proved that constraint dependency grammars were PAC-learnable, making them the first grammatical formalism with strictly greater power than context-free grammars to be shown to be learnable. Again, a big collective yawn. Who cares? The whole NLP world was just too busy trying to induce tri-grams and context-free grammars that they didn't pay any attention. Around 1995 or so, after years of enduring paper rejection after paper rejection, marginallization, and the drying up of research funds, the writing on the wall became quite obvious--the research community just wasn't interested in constraint-based dependency grammars! I dropped out of grad school, went to into VLSI routing, and went on to write several routers which routed the Itanium, Pentium III, IV, and Centrino, and several ARM cores at Intel :-) But I always wondered what might have been.... ...well, it looks like what might have been has actually come true! Menzel & his collegues wrote a complete grammar for German as a constraint grammar, which was very influential. Duchier also picked up the contraint grammar formalism and extended it in interesting ways. Fifteen years after Maruyama published his paper, and ten years after I gave the field up for dead, it has suddenly blossomed and spawned an amazing amount of very creative and productive research! This volume, being the proceedings of a workshop on constraint solving and language processing, shows how this field is absolutely shining with promise and shining with the brillience of the researchers who contribute to the papers gathered herein. Every paper is a gem, but my favorites were: "An Abductive Treatment of Long-Distannce Dependencies in CHR" by Veronica Dahl. What is interesting here is that Dahl has fused Constraint handling Rules (CHR) from the forefront of the constraint satisfaction community's research with constraint-based parsing, to create a very interesting new parsing system. "Problems of Inducing Large-Coverage Connstraint-based Dependency Grammars for Czech" by Ondrej Bojar. So-called "crossover constraints," which rule out crossing dependencies, are very useful for grammers like English which have a strict word order. We always hypothesized that constraint based grammars would be useful even for free-word order languages, but apparently not being able to use these crossover constraints presents some problems. The author gives results with Czech, which shows that there is yet more intreseting work to be done in adapting constraint-based formalism for free-word order languges. "Parsing unrestricted German Text with Defeasible Constraints", by Foth, Daum, and Wolfgang Menzel. Menzel was an early visionary in constraint-based grammar research, and this paper presents very interesting techniques for "relaxing" constraints for parsing eliptical and fragmentary sentences. The WCDG formalism is an extremely promising enhancment to CDG which seems to do a good job of parsing language as its actually spoken--sometimes ungrammatical, sometimes fragmentary. All-in-all, its very gratifying to see that, 15 years on, Maruyama's initial insights into the usefulness of grammatical formalism based on constraints over dependency graphs is blossoming and bearing fruit. Comment | | (Report this)