Alexander Clark (aclark@galactica.it)
Thu, 06 Feb 1997 14:29:25 +0100
At 09:29 04/02/97 GMT, you wrote: > >> >Well, by changing the values of roles in qfpsoas to content-objects >(nom-obj, etc.) we can get the anchoring information we need directly at >the sentence node without any additional principles, assuming that the >CONTENT of the mother is that of its adjunct daughter. I don't know what >other problems might arise under this approach, but it seems reasonable. > >Tony Davis >Cycorp, Inc. >Austin, TX There is a well known philosophical distinction normally credited to Keith Donnellan between attributive and referential uses of NPs that in some sense seems to parallel this distinction between using indexes and content-objects as the values of roles. The example he gives is (1) 'Smith's murderer is insane' In the attributive use - corresponding to using content-objects - we don't know who killed Smith - we have merely found Smith's body and from the manner of the killing we say (1). We could in this case add 'whoever he or she is' parenthetically after the NP The referential use would be used when Jones has been arrested and charged with the crime. He acts very strangely during the trial. Based on this we utter (1) possibly adding (ie Jones) after the NP. It seems that in order to capture this distinction one might need to be able to switch between using the extension of the NP (the idx more or less) and the intension (the nom-obj) as the value of the role. ---------------------------------------------- - Alexander Clark - - alexander_clark@iol.it - ----------------------------------------------
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:36:04 PST