Lexical Heads, Phrase Structure and the Induction of Grammar

Acquiring linguistically plausible phrase-structure grammars from ordinary text has proven difficult for standard induction techniques, and researchers have turned to supervised training from bracketed corpora. We examine why previous approaches have failed to acquire desired grammars, concentrating our analysis on the inside-outside algorithm (Baker, 1979), and propose that with a representation of phrase structure centered on head relations such supervision may not be necessary.