Christopher Manning (cmanning@mail.usyd.edu.au)
Thu, 10 Jul 1997 10:05:02 +1000
On 9 July 1997, Valia Kordoni wrote: > Theodora Alexopolou writes: > > I am interested in work done on Clitic Doubling and Clitic Left > > Dislocation within HPSG. > There is Paola Monachesi's dissertation on clitics in Italian: "A > grammar of Italian Clitics". Tilburg University, which is available > from her web page (http://www.sfs.nphil.uni-tuebingen.de/~paola/). Unfortunately, I don't think Paola's work actually deals with either of these phenomena, and I'm not aware of any work in HPSG that does. I think it is fairly straightforward to see how c.1995 accounts of clitics in HPSG could be extended to doubling (that is accounts in terms of `argument composition' -- i.e., division categories -- implemented via lexical rules). One would propose a rule such as this: [ verb ] [ COMPS L1 + <[1]> + L2 ] [ COMPS L1 + <[1] [CASE dat]> + L2 ] ==> [ CLTS W U {[1]} ] [ CLTS W ] (There is a minor computational point to be careful of here, since such a rule could reapply to its own output ad infinitum, but since CLTS is a set, this reapplication would be harmless.) It is actually a little more complicated, I think, to rework 1997 HPSG accounts (Sag & Miller 1997, Abeille, Godard & Sag 1997) where cliticization is done in a sortal hierarchy which divides elements of ARG-ST into affixes and non-affixes (leaving out a few complications). It seems that you would have to propose a three way division of synsems into pure-affix, aff-canon and canon, with only pure-affixes forbidden from appearing in valence lists, and aff-canons leading a double life as affixes and as members of valence lists. Chris Manning
This archive was generated by hypermail 2.0b3 on Fri Dec 18 1998 - 20:36:08 PST