Pieter Theron, ISSCO (University of Geneva)
All natural language processing systems require language knowledge coded in a computationally tractable way. To hand-code this knowledge is expensive and time-consuming. An alternative approach is to attempt the automatic acquisition of such knowledge. This presentation will introduce our research on a language-independent method for the automatic acquisition of morphological rules. The two-level model of morphology [Koskenniemi,1983] serves as the target coding formalism. Apart from its practical advantages, our automatic acquisition method is also interesting from a linguistic point of view: The acquired set of rules can be viewed as an optimal model of the sound changes which occur during morphological transformations on a given set of words. The accuracy of this acquired model can be empirically verified on previously unseen words.
Pius ten Hacken, University of Basle
Reusability has been a big issue in NLP since approximately the mid-1980s. It raises various questions, one of which is what is the extent of the domain meant to be reusable. Since the mapping between text words and lexemes is a task common to all text-based NLP-systems and independent of the application and the theory of a particular NLP-system, it is an ideal domain for reusability. A lexeme is an inflectional paradigm with a name. Word Manager (WM) covers this domain entirely. It consists of rules describing the systems of inflection and word formation of a language, rules for splitting up a text word into two word forms (clitics) or combining text words (multi-word units), and an object-oriented lexical database system. More information on WM is available at the following URL:
Cornelia Tschichold, LTLP Univ. of Neuchatel (& English Seminar, Basel)
The LTLP has developed a prototype of a new grammar checker specifically geared to the needs of French speakers writing in English. Most commercial grammar checkers on the market today are meant to be used by native speakers of a language who have good intuitions about their own language competence. Non-native speakers of a language, however, have different intuitions and are very easily confused by false alarms, i.e. error messages given by the grammar checker when there is in fact no error in the text. In our project, we concentrated on building a grammar checker that keeps the rate of over-flagging down and on developing a user-friendly writing environment which contains, among other things, a series of on-line helps. The grammar checking component uses island processing (or chunking) rather than a full parse. This approach is both rapid and appropriate when a text contains many errors. We explain how we use automata to identify multi-word units, detect errors (which we first isolated in a corpus of errors) and interact with the user.
Margaret King, ISSCO
I'm presenting ISSCO's activities at the Zurich meeting. Could I get out of sending you a summary by suggesting that anyone who wants to check us out beforehand should look at ISCCO's activities
Jean-Luc Cochard, IDIAP
Martin Rajman, EPFL Articicial Intelligence Lab
Eric Keller, Université de Lausanne
Paraic Sheridan, ETH Zurich
Beat Pfister et al., TIK, ETH Zurich
Norbert E. Fuchs, Department of Computer Science, University of Zurich
Cathy Berthouzoz, LATL, Université de Genève
Martin Volk, Dept. of Computer Science, University of Zurich