nmod:desc

home edit page issue tracker

This page pertains to UD version 2.

It appears that you have Javascript disabled. Please consider enabling Javascript for this page to see the visualizations.

`nmod:desc`: descriptor modifier in nominal

** IMPLEMENTATION IN PROGRESS **

This relation subtype applies to nominal modifiers that we term descriptors. These are bare nominals that occur in or with a name, and are not prepositional/possessive or part of the English compound construction. For personal names, titles and role descriptions are a prime example. Business names may include descriptors like Inc. Semantically, a descriptor assigns a referent to a class or may disambiguate the referent where the main part of a proper name is not a sufficiently unique identifier. Crucially, omitting the descriptor does not affect the grammaticality of the expression (and does not come across as an abbreviation of the full name).

Descriptors are bare modifiers (i.e. the descriptor noun lacks a determiner, possessive, or numeral). This distinguishes them from appositives: the appos relation holds between two full nominals. In principle, descriptors are a special case of nmod:unmarked.

Personal Names

One descriptor construction is the pre- or postmodification of a personal name with a (bare nominal) role or title:

** NOTE: we need to standardize UPOS. SYM says that Dr. and Mr. are NOUN but in English treebanks they are predominantly PROPN. **



  
    input
    brat
    info
  
Mr./NOUN Magoo/PROPN
nmod:desc(Magoo, Mr.)
{
    "text": "Mr. Magoo ",
    "entities": [
        [ "embedded-1-T1", "NOUN", [ [ 0, 3 ] ] ],
        [ "embedded-1-T2", "PROPN", [ [ 4, 9 ] ] ]
    ],
    "relations": [
        [ "embedded-1-R1", "nmod:desc", [ [ "arg1", "embedded-1-T2" ], [ "arg2", "embedded-1-T1" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "1" ]
}



  
    input
    brat
    info
  
Professor/NOUN Charles/PROPN J./PROPN Fillmore/PROPN , Ph.D./PROPN
nmod:desc(Charles, Professor)
flat(Charles, J.)
flat(Charles, Fillmore)
nmod:desc(Charles, Ph.D.)
{
    "text": "Professor Charles J. Fillmore , Ph.D. ",
    "entities": [
        [ "embedded-2-T1", "NOUN", [ [ 0, 9 ] ] ],
        [ "embedded-2-T2", "PROPN", [ [ 10, 17 ] ] ],
        [ "embedded-2-T3", "PROPN", [ [ 18, 20 ] ] ],
        [ "embedded-2-T4", "PROPN", [ [ 21, 29 ] ] ],
        [ "embedded-2-T5", "token", [ [ 30, 31 ] ] ],
        [ "embedded-2-T6", "PROPN", [ [ 32, 37 ] ] ]
    ],
    "relations": [
        [ "embedded-2-R1", "nmod:desc", [ [ "arg1", "embedded-2-T2" ], [ "arg2", "embedded-2-T1" ] ] ],
        [ "embedded-2-R2", "flat", [ [ "arg1", "embedded-2-T2" ], [ "arg2", "embedded-2-T3" ] ] ],
        [ "embedded-2-R3", "flat", [ [ "arg1", "embedded-2-T2" ], [ "arg2", "embedded-2-T4" ] ] ],
        [ "embedded-2-R4", "nmod:desc", [ [ "arg1", "embedded-2-T2" ], [ "arg2", "embedded-2-T6" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "2" ]
}



  
    input
    brat
    info
  
actor and social activist Martin Sheen
nmod:desc(Martin, actor)
conj(actor, activist)
cc(activist, and)
amod(activist, social)
flat(Martin, Sheen)
{
    "text": "actor and social activist Martin Sheen ",
    "entities": [
        [ "embedded-3-T1", "token", [ [ 0, 5 ] ] ],
        [ "embedded-3-T2", "token", [ [ 6, 9 ] ] ],
        [ "embedded-3-T3", "token", [ [ 10, 16 ] ] ],
        [ "embedded-3-T4", "token", [ [ 17, 25 ] ] ],
        [ "embedded-3-T5", "token", [ [ 26, 32 ] ] ],
        [ "embedded-3-T6", "token", [ [ 33, 38 ] ] ]
    ],
    "relations": [
        [ "embedded-3-R1", "nmod:desc", [ [ "arg1", "embedded-3-T5" ], [ "arg2", "embedded-3-T1" ] ] ],
        [ "embedded-3-R2", "conj", [ [ "arg1", "embedded-3-T1" ], [ "arg2", "embedded-3-T4" ] ] ],
        [ "embedded-3-R3", "cc", [ [ "arg1", "embedded-3-T4" ], [ "arg2", "embedded-3-T2" ] ] ],
        [ "embedded-3-R4", "amod", [ [ "arg1", "embedded-3-T4" ], [ "arg2", "embedded-3-T3" ] ] ],
        [ "embedded-3-R5", "flat", [ [ "arg1", "embedded-3-T5" ], [ "arg2", "embedded-3-T6" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "3" ]
}



  
    input
    brat
    info
  
friend of the pod Ron Klain
nmod(friend, pod)
nmod:desc(Ron, friend)
flat(Ron, Klain)
{
    "text": "friend of the pod Ron Klain ",
    "entities": [
        [ "embedded-4-T1", "token", [ [ 0, 6 ] ] ],
        [ "embedded-4-T2", "token", [ [ 7, 9 ] ] ],
        [ "embedded-4-T3", "token", [ [ 10, 13 ] ] ],
        [ "embedded-4-T4", "token", [ [ 14, 17 ] ] ],
        [ "embedded-4-T5", "token", [ [ 18, 21 ] ] ],
        [ "embedded-4-T6", "token", [ [ 22, 27 ] ] ]
    ],
    "relations": [
        [ "embedded-4-R1", "nmod", [ [ "arg1", "embedded-4-T1" ], [ "arg2", "embedded-4-T4" ] ] ],
        [ "embedded-4-R2", "nmod:desc", [ [ "arg1", "embedded-4-T5" ], [ "arg2", "embedded-4-T1" ] ] ],
        [ "embedded-4-R3", "flat", [ [ "arg1", "embedded-4-T5" ], [ "arg2", "embedded-4-T6" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "4" ]
}

A descriptor must be omissible, but not all omissible parts of a name are descriptors: if the main part of a name has multiple parts such that there are at least two plausible heads, flat is the appropriate choice, as is the case with the main parts of a personal name.

Non-numeric generational suffixes are also nmod:desc (but see Numbered Entities below):



  
    input
    brat
    info
  
JFK/PROPN Jr./PROPN
nmod:desc(JFK, Jr.)
{
    "text": "JFK Jr. ",
    "entities": [
        [ "embedded-5-T1", "PROPN", [ [ 0, 3 ] ] ],
        [ "embedded-5-T2", "PROPN", [ [ 4, 7 ] ] ]
    ],
    "relations": [
        [ "embedded-5-R1", "nmod:desc", [ [ "arg1", "embedded-5-T1" ], [ "arg2", "embedded-5-T2" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "5" ]
}

Business Names

Corporate suffixes like Inc., Corp., and LLC (as well as their non-abbreviated forms) are considered optional descriptors as they are often omitted from the name of the company. Regardless of etymology, they are tagged PROPN.



  
    input
    brat
    info
  
Apple/PROPN Inc./PROPN
nmod:desc(Apple, Inc.)
{
    "text": "Apple Inc. ",
    "entities": [
        [ "embedded-6-T1", "PROPN", [ [ 0, 5 ] ] ],
        [ "embedded-6-T2", "PROPN", [ [ 6, 10 ] ] ]
    ],
    "relations": [
        [ "embedded-6-R1", "nmod:desc", [ [ "arg1", "embedded-6-T1" ], [ "arg2", "embedded-6-T2" ] ] ]
    ],
    "attributes": [],
    "error": false,
    "sentlabels": [ "6" ]
}

Complex Dates and Addresses

These are considered to have time or place modifiers, and thus fall under nmod:unmarked.

Idiosyncratic Name Patterns Analyzed as Flat Expressions

In addition to the main parts of a personal name, further name patterns analyzed with flat rather than a headed construction are given below. Note that the morphosyntax of name patterns differs by language, with some exhibiting agreement (as a sign of head or modifier status) where it is lacking in English, so a flat analysis of an English expression may not hold for its translation into another language.

Numbered Entities

When the name of a category combines with a (pseudo)number to form a name, and the number is serving as an identifier rather than as a quantifier, we use flat:¹

page/NOUN 394/NUM
flat(page, 394)

Route/PROPN 66/NUM
flat(Route, 66)

Room/PROPN B/PROPN
flat(Room, B)

The “number X” construction also serves to clarify that a number is serving as an identifier; “number X” forms a flat expression that can be combined within a larger flat expression:

What is behind door/NOUN number/NOUN 3/NUM ?
flat(door, number)
flat(number, 3)

A product name that includes a model or version number is considered a flat expression:

Nikon/PROPN D7000/PROPN
flat(Nikon, D7000)

Firefox/PROPN 43/NUM
flat(Firefox, 43)

World/PROPN War/PROPN II/NUM
flat(World, War)
flat(World, II)

For simplicity, regnal/generational number suffixes are considered part of flat expressions just like version numbers, though they semantically resemble Jr. and Sr.:

Queen/PROPN Elizabeth/PROPN II/NUM
nmod:desc(Elizabeth, Queen)
flat(Elizabeth, II)

Toponyms

We also use flat for toponyms where the noun representing the entity type comes first. This word order is anomalous relative to other English constructions such that tests for headedness are inconclusive:

Lake/PROPN Mead/PROPN
flat(Lake, Mead)

Mount/PROPN Everest/PROPN
flat(Mount, Everest)

By contrast, if the identificational part of the toponym precedes the entity type, this is analyzed as a regular right-headed compound dependency:

Mirror/PROPN Lake/PROPN
compound(Lake, Mirror)

This is a departure from a guideline articulated previously: according to de Marneffe et al. (2021), p. 285, the (pseudo)number attaches as nmod, though the main point of that passage is to articulate why nummod is not appropriate (no rationale for headedness is presented). ↩

nmod:desc in other languages: [en]