Invited Lecture


Smart vs. Solid Solutions in Computational Linguistics

Roland Hausser
Computational Linguistics
University of Erlangen-Nuremberg
Germany

A smart solution tries to solve a particular problem with a minimum of effort and a maximum of effect by avoiding the difficult aspects of the task at hand. There are many applications in which smart solutions provide the user with a partial, yet highly welcome improvement. However, when turning to a slightly different application, the smart solution usually cannot be reused, leading to a multiplication of effort. Furthermore, it is difficult if not impossible to increase accuracy.

A solid solution, in contrast, is based on ready-mady off-the-shelve components such as complete on-line dictionaries, morphological, syntactic, and semantic parsers and generators for different languages, annotated reference and monitor corpora for different domains, etc. These components are developed and maintained in the context of basic research, but made available to the full range of practical applications.

Given that linguistics, like any science, is confronted with problems which are still theoretically unsolved, both smart and solid solutions have their place. This talk will describe their possible interaction in different applications in order to simultaneously reduce long-term cost, raise overall quality, and support theoretical development.