Machine learning theory and practice as a source of insightinto universal grammarWe gratefully acknowledge Michael Collinsʼs aid in drafting section 5 of this paper. Antecedents of this paper were presented at the Workshop on Computational Linguistics, University College, London, 2004; Computational Linguistics in the Netherlands 2004, Leiden; the Linguistics Colloquium, University of Illinois at Urbana-Champaign, 2005; Construction of Meaning: Stanford Semantics and Pragmatics Workshop, Stanford University, 2005; the 2005 LSA Summer Institute, Cambridge, MA; ESSLI 2005, Edinburgh; the Computer Science Colloquium, University of Essex, 2006; the Cognitive Science and Linguistics Colloquium, Carleton University, Ottawa, 2006; and the Linguistics Department Colloquium, Stanford University, 2006. The first author presented some of the proposals contained in this paper in his Ph.D Research Seminar in the Philosophy Department, Kingʼs College, London, during the second semester, 2006. We are grateful to the participants of these forums and to three anonymous reviewers for the Journal of Linguistics for much useful feedback, some of which has led to significant modifications in our proposals. We would also like to thank Joan Bresnan, Alex Clark, Eve Clark, Jennifer Cole, Jonathan Ginzburg, Ray Jackendoff, Dan Jurafsky, Ruth Kempson, Chris Manning, Fernando Pereira, Steve Pinker, Geoff Pullum, Ivan Sag, Richard Samuels, Barbara Scholz, Gabriel Segal, Richard Sproat, and Charles Yang for helpful discussion of many of the ideas proposed in this paper. We are indebted to the Stanford Symbolic Systems Program, Ivan Sag, and Todd Davies for providing a welcoming venue for completion of the paper. The first author is grateful to the Leverhulme Foundation for supporting his work there. The first authorʼs research was supported by grant number RES-000-23-0065 of the Economic and Social Research Council of the United Kingdom. Needless to say, we bear sole responsibility for these ideas.

Показать метаданные