Induction of Linguistic Knowledge

Machine Learning Techniques for Language Engineering and Linguistics

The growing availability of language corpora combined with large computing power has opened up exciting new possibilities for computer-aided language processing. ILK, Induction of Linguistic Knowledge, is a research programme in which inductive learning algorithms are developed and employed in solving natural language problems. The research is intended to be useful in two areas:
  1. Language Engineering. In language engineering, building applications is often hampered by the knowledge acquisition bottleneck: it is hard to make linguistic knowledge explicit `by hand', and to make it reusable for different language tasks or languages. The application of inductive learning holds great promise for alleviating this bottleneck. It has been shown that the application of inductive learning results in more accurate and robust systems than those developed by hand. Moreover, the learning methods are language-independent. Automatically learned models can be put into direct practice as modules in language and speech engineering applications.

  2. Linguistic and Cognitive Modelling. As a new method in linguistic analysis, inductive learning allows us to discover linguistic knowledge in corpora, dictionaries, and sets of language examples. They also provide a new perspective on old philosophical and cognitive science problems concerning the representation and acquisition of linguistic knowledge.
The research programme, co-directed by dr Walter Daelemans and dr Antal van den Bosch, is part of CLS (Center for Language Studies, Faculty of Arts, Tilburg University), and is sponsored by NWO (Netherlands Organization for Scientific Research, Foundation for Language, Speech, and Logic), SoBU (Cooperation Association Tilburg and Eindhoven Universities), and STW (Foundation for the Technical Sciences, The Netherlands).

Last update: 24 January 2000