GOLD 2010

The current hierarchical organization of Numeral as compared to Determiner is problematic:

- Numeral is-a Quantifier (this is correct with respect to the semantic characteristics of Quantifiers and Numerals)
- Quantifier is-a Determiner (this is correct with respect to the *syntactic* characteristics of quantifying determiners)

However, in the following sentences, "three" is hardly a Determiner, but rather, the head of an NP. According to its semantics, however, it is clearly a Numeral:

(1) There were *three* of them.

(2) Wir sind die *drei* von der Tankstelle. (German)
we are the *three* from the gas.station

One may consider to promote Quantifier to a top-level category in order to distangle semantically-determined parts of speech (e.g., quantifier, numeral) from syntactically-determined parts of speech (e.g., determiner).

References:
In existing annotation schemes for parts of speech, there normally is no category "Quantifier", but rather, Quantifiers are grouped together with "indefinite determiner" (for quantifying determiners, e.g., "some", "any") or "indefinite pronouns" (for [pro]nominal quantifiers, e.g., "someone", "anybody"), cf. Wilson and Leech (1996, www.ilc.cnr.it/EAGLES/pub/eagles/lexicons/elm_en.ps.gz). An explicit Quantifier concept entails the above-mentioned problems.

The rdfs:comment on gold:orderingRelation states it to be "Any relation that establishes a linear ordering of linguistic units", yet neither its rdfs:domain nor its rdfs:range is restricted to gold:LinguisticUnit. This is either misleading if "linguistic unit" in the comment does not refer to gold:LinguisticUnit or the respective restrictions are missing, or do I miss something?

Is there a language in which Processive morphemes are found? If so, an example would be helpful. If not, this concept should be deleted.

This should be subclassified into SpokenUnit and WrittenUnit. The current children of FormUnit should then become the children of the correct subconcept. (Grapheme should become the child of WrittenUnit, the rest children of SpokenUnit.)

Is proverb a genuinely distinct part-of-speech or is it a function of syntax?

As the sister nodes of DiscourseUnit (FormUnit, GrammarUnit, SemanticUnit) all have child nodes with subtypes, it would make sense to have subtypes for DiscourseUnit as well, e.g. 'Narrative Unit' (e.g., Abstract, Coda, etc.), 'Conversational Unit' (e.g., turn, opening, closing, etc.).

Apparently there exist alternative definitions of "essive" in Finno-Ugric languages:

The Hungarian "formativus, or essivus-formalis '-ként' ... usually expresses a position, task and manner of the person or the thing." (Nose 2003)

"Haspelmath & Buchholz (1998:321) explained the function of the essive case as 'role phrases'. Role phrases represent the role of the function in which a participant appears. They regard the role phrases as adverbial."
(Nose 2003, p. 117)

In Estonian the essive case means such things as `(I played golf)
as a student', `(I worked) as a bartender', `(you look) tired',
`(he's very good) as a dancing partner', `(we parted) as friends'.
This doesn't sound like the definition you quoted, but is similar
(though not identical) to the meaning of the Hungarian form.
(Ivan A. Derzhanski, email 2010/06/15)

Apparently, the GOLD community relies on different definitions
of essive (referring to Lyons 1968: 299, 301; Crystal 1985: 112;
Blake 2001, see http://linguistics-ontology.org/gold/2009/EssiveCase)

The definition there defines the term as often used in descriptions of Daghestanian languages, where an essive is a kind of locative. It is thought that the Finnish/Estonian essive started life that way too, but in the contemporary languages _karhu-na_, given on that page as an example and glossed `at the bear', means rather `as a bear', `in one's being a bear'. ... The Hungarian form on the other hand was never a locative, as far as I know.
(Ivan A. Derzhanski, email 2010/06/15)

References:
Masahiko Nose (2003), Adverbial Usage of the Hungarian Essive Case
personal communication with Ivan A. Derzhanski and Csaba Oravecz

Aikhenvald not only lists Noun, Nominal and Numeral Classifiers, but also Verbal Classifiers, Locative & Deictic Classifiers and Classifiers in PossessiveConstructions. These types should be added as children of the concept Classifier.

This seems to overlap with Copula. We will need to investigate whether we really need Copulative as a distinct concept from Copula.

It is wrong to classify copula as a child of verbal. There are non-verbal copula such as pronoun and particle copula as well as zero copula as shown in the examples. Hence, this classification should be reconsidered.

Changed 'verb' in definition to elements

Also noticed the source for this is Stassen 2008 - We have Stassen 2005 in the bibliography but not 2008. Is the 2005 source correct?
-Matt

It should be reconsidered whether Adjectival should be a child of Predicator, because this seems to be theory-specific and not general. Adjectives exhibit very diverse behaviors in the world's languages.

The concept "LexicalItem" refers to "dictionary", which is not defined in GOLD, and it does not mention Lexicon at all, though a Lexicon is defined as a set of LexicalItems.

Suggestion: replace "dictionary" in the definition by "Lexicon"

Is this concept really needed? Do any languages use this?

Is there a language in which this is found? If so, an example would be helpful.

This concept needs a better definition which explains the term in a better way and also why the concept is necessary.

This concept needs a better definition which explains the term in a better way and also why the concept is necessary.

This concept needs a better definition which explains the term in a better way and also why the concept is necessary.

This concept needs a better definition which explains the term in a better way and also why the concept is necessary.

It is not clear whether Retracted is a LabialProperty. From the literature it does not appear to be. A different parent should be considered, e.g. MannerProperty.

Blackfoot example of plural number and animate gender.
2009-06-04 13:28:09
man- me man- mak
III- food VEG- good
good food

References:
Mithun (1999:99) FROM Frantz (1991:8-9)

Comment: the example is from Bininj Gunwok but the references refer to other languages.

References:
III- food VEG- good
good food

References:
Mithun (1999:99) FROM Frantz (1991:8-9)

Comment: the example is from Bininj Gunwok but the references refer to other languages.

The current definition of CaseProperty (with reference to http://www.grammaticalfeatures.net/features/case.html) seems to apply to both "grammatical case" (e.g., ErgativeCase) and "semantic case" (e.g., MalefactiveCase). Although this can be justified given the usage of "case" in the literature (e.g., the traditional, morphosyntactic concept of case vs. Fillmore's case roles 1966), this design decision conceals a difference that is crucial to modern annotation schemes used for the development of NLP tools. If GOLD is intended to be suitable for corpus linguists, computational linguists and NLP engineers, as well, this difference should be made explicit, e.g., by distinguishing GrammaticalCaseProperty from SemanticCaseProperty.

References:
The suggested differentiation between grammatical case and "non-grammatical case" was previously implemented in the TDS Ontology (http://languagelink.let.uu.nl/tds/main.html).
In annotation schemes for inflectional languages, e.g., German, grammatical case is modelled as an aspect of morphology (e.g., Brants et al. 2005, http://www.springerlink.com/content/khv5702313320560/), whereas semantic roles are modelled as an aspect of frame annotation (e.g., Burchard et al. 2006, http://www.coli.uni-saarland.de/%7Epado/pub/papers/lrec06_burchardt1.pdf).