Abbreviations

For more information about the lecture halls and seminar rooms, see our lecture room page. The names listed under "Technical Assistance" are student volunteers who will act as a contact person for technical questions of the lecturers and workshop speakers during the course or workshop.

Grammar induction and language evolution

Recently, much progress has been made in developing methods for learning grammars from natural language text. The first part of the course gives an overview of such algorithms, starting with those which move through a space of context-free grammars using nonterminal chunking, merging and/or splitting. We discuss several local search heuristics, as well as some global objective functions (Bayesian posterior probability, maximum likelihood) and inference procedures (e.g. EM). We evaluate strengths and weaknesses of these models on several metrics, and consider some extensions that work with richer input data (phonological phrasing, semantics) and different grammatical formalisms (dependency grammar, categorial grammar, TAG, DOP). In the second part of the course, we discuss applications of these algorithms in models of language acquisition, change and evolution. We show how the constraints on the search space necessary for successful learning emerge automatically in iterated learning. We consider biological evolution of the inductive bias, and look at the nativist-empiricist controversy from this perspective.