Putting PNG features on a pron_rel


For constructions where the possessor pronoun is an affix on the possessum, I need there to be at minimum two sets of PNG features stored on the possessor+possessum form: the possessum’s PNG features, and the possessor’s PNG features. If the possessor pronoun affix does any agreeing with the possessum, then I need three sets of PNG features. Now, I’ve made a spot for agreement features at HEAD.POSS.POSS-AGR, but I’m having problems with the posessor’s PNG features getting mapped to the usual CONT.HOOK.INDEX.PNG path, which is where the possessum’s features should live. I’ve tried to remedy this by having the features map to the ARG0 of the pron_rel that the affix adds, but it turns out that ARG0 is of type individual, while PNG is a feature ofref-ind, so this doesn’t unify. Is there some way around this issue?

Here’s my pron_rel and some of the path to it, for reference:

LKEYS.ALTKEYREL #altkeyrel & noun-relation &                                                                   
                   [ PRED "pron_rel",                                                                       
                     LBL #lbl2,                                                                             
                     ARG0 #possessor & [ COG-ST activ-or-more,                                              
                                         SPECI + ] ] ]

And then the attempt at putting features on ARG0 that didn’t work:
LKEYS.ALTKEYREL #altkeyrel & noun-relation &
[ PRED "pron_rel",
LBL #lbl2,
ARG0 ref-ind & #possessor &
[ PNG [ NUM sg,
PER 1st ],
COG-ST activ-or-more,
SPECI + ] ] ] ].


Further complication: I realized I had written this rule as though it were a lexical item rather than a lexical rule, so I don’t think I can actually even use the LKEYS.ALTKEYREL to identify the pron_rel – I’m pretty sure the pron_rel should just be one on a list of several rels in the C-CONT. So I need to figure out that as well.


The pron_rel should come in via C-CONT.RELS. ref-ind is a subtype of individual, so I don’t see why that should fail to unify.

Also — I don’t understand why you’d ever need three sets of PNG features. First, I’d be a little surprised if you ever found possessors marked as affixes on possessums that also agree with the possessum. Is that really attested? Second, if you did, then those affixes, being affixes on the possessum, could directly talk about its INDEX.PNG.


Right, I can definitely just do the two sets PNG features, then. :slight_smile:

Is there anything like LKEYS.KEYREL and LKEYS.ALTKEYREL for lexical rules that allows you to point at specific things on the RELS list? From what I can see, LKEYS is lexical-item-specific. Or should I just be trying to add features to one of the items on the C-CONT.RELS list using RELS.LAST…etc syntax?


There is no equivalent to LKEYS on lex rules — probably C-CONT.RELS.LIST.FIRST is what you want :slight_smile:


An update on this (post advisee meeting :slight_smile: ). So the tricky thing was that the PNG features are added in features.py, while the rest of the lexical rule is added in my library. So I had to be able to talk about something on the RELS list that I was adding elsewhere. Turns out, tdl.py won’t merge the two different ways of talking about diff-lists: if you use the RELS.LIST syntax in one place, you have to use it in both. So instead, I made features.py add the PNG features at the path C-CONT.RELS <! [ ARG0.PNG, and then made sure it tacked on a ]!> later in features.py.