In this presentation we describe the Emile grammar induction approach. The Emile approach is based on notions from categorial grammar, which is known to generate the class of context-free languages. Emile learns from positive examples only. We describe the algorithms underlying the approach and present some interesting practical results on small and large text collections. The experiments show that Emile
might  be a valuable tool for syntactical and semantical analysis of large text corpora. A very promising observation is that Emile already starts of
converge on datasets of moderate size like the bible.

Marten Trautwein (Marten.Trautwein@ps.net)
        Perot Systems Nederland BV
        P.O.Box 2729,
        NL-3800 GG Amersfoort,
        The Netherlands

Marco Vervoort (vervoort@wins.uva.nl)
        University of Amsterdam, FdNWI
        Plantage Muidergracht 24,
        NL-1018 TV Amsterdam,
        The Netherlands