Path: utzoo!utgpu!news-server.csri.toronto.edu!bonnie.concordia.ca!uunet!mcsun!ukc!canon!wachtel From: wachtel@canon.co.uk (Tom Wachtel) Newsgroups: comp.lang.prolog Subject: Re: Efficiency of DCGs and chart parsers Message-ID: <1991Jan28.131922.727@canon.co.uk> Date: 28 Jan 91 13:19:22 GMT References: <8014@castle.ed.ac.uk> <1549@gufalet.let.rug.nl> Organization: Canon Research Europe, Guildford, UK Lines: 13 gosse@gufal02.let.rug.nl (Gosse Bouma) writes: >I find that reasonable, given the fact that >feature-unification, rather than prolog-unification, has to be performed The contrast between feature-unification and prolog-unification is unnecessary and expensive. Mapping high-level feature matrix notation to prolog terms in a controlled way allows you to treat feature-unification as prolog-unification, directly on the whole feature matrix represented as a prolog term with a known structure. It becomes the same thing. Tom Wachtel (wachtel@canon.co.uk)