I'm wondering how open-class families are intended to be used in OpenCCG.
The 'Rough Guide' says:
"Note that a closed family does not exactly correspond to the notion of a closed class word, as open class words (especially verbs) are often listed as members of closed families, in order to assign them appropriate subcategorization frames." (p.6)
However, in practice it seems that open class words are always listed as member of closed class families, at least based on the sample grammars included in the distribution.
It appears that the <member-of> tag is only used inside the dict.xml file, and that the transformations that apply to this take all stems defined as <member-of family="X"> and place them inside a <member> tag in the lexicon, with the family being defined as closed.
In effect, closed families seem to be treated as the only kind of family, in order to correctly associate a stem with its category type.
How would one associated entries in the morph.xml file if a family were listed as an open type?
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Using open versus closed class items was a big part of the grammar that Julia, Gann and I hooked up to OpenCCG (then Grok) based on the lexicon that Julia created from the Penn Treebank. See the following paper:
I'm wondering how open-class families are intended to be used in OpenCCG.
The 'Rough Guide' says:
"Note that a closed family does not exactly correspond to the notion of a closed class word, as open class words (especially verbs) are often listed as members of closed families, in order to assign them appropriate subcategorization frames." (p.6)
However, in practice it seems that open class words are always listed as member of closed class families, at least based on the sample grammars included in the distribution.
It appears that the <member-of> tag is only used inside the dict.xml file, and that the transformations that apply to this take all stems defined as <member-of family="X"> and place them inside a <member> tag in the lexicon, with the family being defined as closed.
In effect, closed families seem to be treated as the only kind of family, in order to correctly associate a stem with its category type.
How would one associated entries in the morph.xml file if a family were listed as an open type?
Names (proper nouns), numbers, dates, etc. are well-suited to open families. The association is via POS.
Using open versus closed class items was a big part of the grammar that Julia, Gann and I hooked up to OpenCCG (then Grok) based on the lexicon that Julia created from the Penn Treebank. See the following paper:
http://www.ircs.upenn.edu/~juliahr/Papers/JLAC/jlac.ps
BTW, for what it's worth, there is a bunch of CCG related information, including a number of publications, on the following page:
http://groups.inf.ed.ac.uk/ccg/
Jason