Hans Uszkoreit (hansu@coli.uni-sb.de)
Thu, 25 Jul 1996 18:24:07 +0200 (MET DST)
Shuly Wintner asks a very good question: What exactly is an HPSG? How can we decide whether a given grammar is an HPSG or not? Since the underlying mathematical formalism has Turing-machine power, are there any reasonable and truly effective constraints on "true HPSGs"? One might further ask: How come that Kasper, Kiefer, Netter, Vijay-Shanker were able to show that HPSGs can be compiled into TAGs if HPSG is of type 0? We can observe in the actual grammar development work in the HPSG framework that HPSG is becoming less PS and maybe even less H. What then is the G? I think because of the nature of HPSG Shuly's question does not have an answer. I view it as an advantage of HPSG that the framework actually is a construction kit for formalized concrete grammar models. For each of these models, Shuly's questions should have appropriate answers. When I saw Shuly's question, I decided to post some notes that I had written earlier on the view of HPSG as a construction kit. You may find them at http://www.coli.uni-sb.de/~hansu/hpsg.html I would be grateful for comments. I would also be happy to supply examples from actual HPSG work in computational linguistics if you have doubts about my view of the essence of HPSG. Hans Uszkoreit
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:33:28 PST