From discussions with Laurence Anthony at CL'17, it was suggested that a further out-of-band storage format be included.
Essentially, the idea is to have a dedicated meta/ sub path in formats that support it to include metadata files for files and associate them by name. It is not intended that this be used for 'exported' data, but instead be a suggestion for implementors to use this format internally when they're designing their own internal storage, as a kind of 'official' out-of-band storage schema.
Using the following scheme vaguely according to AntCorGen's corpus storage format:
corpusName/
abstract/
a.txt
b.txt
introduction/
a.txt
b.txt
id/
a.txt
b.txt
meta/
a.js <-- Metadata file, as per the normal .js format
b.js
group.js <-- arbitrary common fields referenced by the 'group' field in {a,b}.js as a file URI
conclusion/
a.txt
b.txt
Suggested additions include the option to have the group or root top level field support file:// URIs so that common features can be put in one place - see #3
From discussions with Laurence Anthony at CL'17, it was suggested that a further out-of-band storage format be included.
Essentially, the idea is to have a dedicated meta/ sub path in formats that support it to include metadata files for files and associate them by name. It is not intended that this be used for 'exported' data, but instead be a suggestion for implementors to use this format internally when they're designing their own internal storage, as a kind of 'official' out-of-band storage schema.
Using the following scheme vaguely according to AntCorGen's corpus storage format:
Suggested additions include the option to have the
grouporroottop level field supportfile://URIs so that common features can be put in one place - see #3