Re: Comments and questions on the specification

Hello Sander,

Thanks for your interest in the model.

On Wed, Mar 1, 2017 at 11:00 AM, Sander Stolk <ssstolk@gmail.com> wrote:

> Dear lemon community,
>
> For my PhD research (working with historical thesauri), I've been looking
> into the lemon modules and what they can express. There are a few comments
> I'd like to place on the current specification/report. The majority of
> these are simply pointing out some minor mistakes or inconsistencies in the
> documentation, others are matters that are not wholly clear to me as a
> reader (and therefore possibly to others as well) and may require
> additional clarification.
>
> The sections mentioned below refer to those in the current specification (
> http://www.w3.org/2016/05/ontolex/).
>
>    - *section 2.2*
>    The namespace  <http://www.w3.org/ns/lemon/all>   currently provides
>    module ontolex only, instead of all modules. Can this be remedied?
>
> It is supposed to return this document:

https://github.com/cimiano/ontolex/blob/master/Ontologies/all.owl

Philipp> Could you update this?

>
>    - *section 5.2*
>    decomp:constituent and rdf:_[number] are currently both required to
>    state the order of constituents. This is, to me, a logical choice,
>    considering one would not want to create multiple subproperties for
>    decomp:constituent just to indicate the ranking of the constituent (e.g.
>    decomp:constituent_1, decomp:constituent_2, etc). Considering rdf:_1,
>    rdf:_2 etc. are subproperties of rdfs:member, however, I would still have
>    expected decomp:constituent to at least be asserted as subproperty of
>    rdfs:member. This would make the relation between decomp:constituent
>    and any required sequence a lot more apparent.
>
> That seems like a good suggestion.

>
>    - *section 6.1*
>    One of the example figures contains a mistake. See the image below the
>    text "The following example gives an example of a sense relation:". Here,
>    the SenseRelation is displayed between references rather than between
>    LexicalSenses. (The Turtle below is correct, however.)
>
> Yes, the fixed image is here:

https://github.com/cimiano/ontolex/blob/master/Examples/vartrans/example2.png


>
>    - *section 9.1*
>    The example (both the figure and in Turtle) speaks of Sense /
>    ontolex:Sense. These URIs do not exist in the final specification. I assume
>    it should read vartrans:SenseRelation instead.
>
> Indeed it should be.

>
>    - *overall*
>    Names for properties, when shown in definition boxes, are sometimes
>    written with an initial capital and sometimes not. This is a minor styling
>    issue, although for readability it would be beneficial to not use initial
>    capital for properties.
>
> I am not sure I see where exactly you are referring to. If you have any
minor corrections, the best way to suggest them is by making a pull request
or an issue to this document

https://github.com/cimiano/ontolex/blob/gh-pages/specification.html

>
>    - *sections 3.6 and 9*
>    These sections discuss how Wordnet and such can be represented in
>    ontolex. The use of LexicalConcept for what is known as a synset in Wordnet
>    is clear. I suspect that it is the intention that the reverse is also true:
>    that a LexicalConcept with the entries that evoke it, and the senses that
>    lexicalize it, should always be expressed as a synset in Wordnet.
>
> Yes, 'synset' and 'lexical concept' are equivalent for most practical
purposes

>
>    - In other words, that ontolex:evokes and ontolex:isLexicalizedSenseOf
>    connections with a LexicalConcept are by definition considered
>    (near-)synonyms. Is this assumption correct?
>
> These two properties differ in the domain, evokes is used to connect a
lexical entry to a lexical concept and lexicalizedSense to connect a
lexical sense to a lexical concept.

>
>    - If it is, I miss super properties for these two ontolex properties
>    that are certainly needed in my own use case -- for historical
>    thesauri. Some of these do not capture synonymy relations but still
>    organize words and their senses through meaning/concepts. Hence the need to
>    state, for example, that a LexicalSense is categorized under a certain
>    concept or semantic field (e.g. "milk, v." under the concept of "Farm"),
>    but is not a direct expression/lexicalization of that concept.
>
>
For the representation of wordnets, you should also look into the Global
WordNet Association format, which I helped develop based on OntoLex lemon.
This is one of the best formats for representing wordnets and has been
adopted by the Open Dutch Wordnet (ODWN) among many others.

http://globalwordnet.github.io/schemas/

For example the property
http://globalwordnet.github.io/schemas/wn#domain_topic would represent the
link between 'milk' and 'farm'.

Regards,
John

>
>    -
>
> I look forward to hearing your thoughts on the above.
>
> Yours faithfully,
> --
> Sander Stolk, MSc MA
> PhD researcher at Leiden University, the Netherlands
>