tr35 spec out of date with regard to DTD (grammaticalFeatures)


this is in the supplemental DTD (and in the data!) but not in the spec:






Mark Davis
December 2, 2020, 7:20 PM

Moving to 39, and capturing as Known Issue on release page.

Steven R. Loomis
November 5, 2020, 8:38 PM

ok, so there’s an implied scope=”everything else” basically if scope is omitted. I think I haven enough to go on.

Steven R. Loomis
October 30, 2020, 9:14 PM

thanks! I did not check that all elements in the DTD were in the spec.

Mark Davis
October 30, 2020, 9:02 PM

Note: for the attributes scope and targets there is currently only one value,



I don’t think we want to put that in the spec, however, since we run the risk of not updating it properly when we add additions.

Mark Davis
October 30, 2020, 8:57 PM

Whoops! Here is the missing information. I think we should add this to the known issues (to unblock you), with a fix version for the spec of 38.1.


The targets contain the specific grammatical entities to which the features apply, such as nominal when they apply to nouns only.

Each of these features can have a scope attribute. The presence of that attribute indicates that the values are limited to a specific subset for certain kinds of entities. For example, a particular language might have an animate gender for nouns, but no units of measurement ever have that case; in another language, the language might have a rich set of grammatical cases, but units are invariant.

The scopes are targeted at messages created for computers, thus a feature may have a narrower scope if for all practical purposes the feature value is not used. For example, it may be possible in theory for a kilogram to be in the vocative case (English poetry might have “O Captain! my Captain!/ our fearful trip is done”, but even in poetry you have little call to write “O kilogram! my kilogram! …”).


  • a scope attribute is only used when there is a corresponding “general” element, one for the same language and target without a scope attribute.

  • the scope attribute values must be narrower (a proper subset, possibly empty) of those in the corresponding general element.





Mark Davis


Steven R. Loomis


Steven R. Loomis

Fix versions