We know that usage can - sometimes - be expressed through relations (notably domain_region/has_domain_region or exemplifies/is_exemplified_by, domain_topic/has_domain_topic).
The first issue of expressing usage through a relation is that it requires identifying proper targets. For instance:
Britain, UK, England, British English ...? for _has_domainregion
British English, British spelling ... ? for exemplifies
law, legal ... ? for _hasdomain topic
etc.
Should we link with the adjective, the noun ('archaism' or 'archaic') ... ?
Beyond geographical locations, I found the following annotations:
'archaic', 'dialectal', 'euphemistic', 'slang', 'literary', 'formal', 'legal', 'vulgar', 'frequent'
When annotating usage, it's highly desirable to have a choice within a predefined set of targets/labels ... that hasn't been defined yet: it's still open. An assessment of the existing set is given here
Second, the relation is to be defined between what and what, in terms of word, sense or synset:
the source
the first term is usually a sense
could it be just a word, whatever the sense? Like exemplifies(f*, taboo) expressing that, whatever the meaning, the word is taboo
can one conceive of a synset exemplifying something ? But a synset can be exemplified by the sense of a word...
the destination
it makes sense to link a sense, for example, with a specific sense of 'formal'
should we link with a synset ? A reasonable assumption given that, if 'derogation' qualifies, so would 'disparagement', 'depreciation', so why not use the synset as target.
The end result would be relations between words, senses and synsets with hybrids like sense to synset ... These are not permissible in the current OEWN
Third, some usage annotations (in Britain they call a fender a `wing´) appear to be ternary if not quaternary relations, not binary ones, but actually result from the composition of same synset membership with a binary relation
similarity of meaning can be captured by same synset membership of 'fender' and 'wing'
with wing has_domain_region Britain in this sense
Not an issue but things are sometimes not that simple.
Some usage notes resist being expressed as relations:
‘Scotch’ is in disfavor with Scottish people and is used primarily outside Scotland
‘continual’ is often used interchangeably with ‘continuous’
in careful usage the noun `enormity´ is not used to express the idea of great size
'Negro' and 'Negroid' are archaic and pejorative today
PR1098 has grouped usage annotations from examples under the same heading 'usage', waiting for sections of definitions to be exported there.
We know that usage can - sometimes - be expressed through relations (notably domain_region/has_domain_region or exemplifies/is_exemplified_by, domain_topic/has_domain_topic).
The first issue of expressing usage through a relation is that it requires identifying proper targets. For instance:
Should we link with the adjective, the noun ('archaism' or 'archaic') ... ?
Beyond geographical locations, I found the following annotations: 'archaic', 'dialectal', 'euphemistic', 'slang', 'literary', 'formal', 'legal', 'vulgar', 'frequent'
When annotating usage, it's highly desirable to have a choice within a predefined set of targets/labels ... that hasn't been defined yet: it's still open. An assessment of the existing set is given here
Second, the relation is to be defined between what and what, in terms of word, sense or synset:
the source
the destination
The end result would be relations between words, senses and synsets with hybrids like sense to synset ... These are not permissible in the current OEWN
Third, some usage annotations (in Britain they call a fender a `wing´) appear to be ternary if not quaternary relations, not binary ones, but actually result from the composition of same synset membership with a binary relation
Some usage notes resist being expressed as relations:
PR1098 has grouped usage annotations from examples under the same heading 'usage', waiting for sections of definitions to be exported there.