Re: relations about lexical entries

Dear John and others,

As someone with a background in linguistic morphology (although it's 
been a while now) and computational lexicography (more recently), I feel 
that the proposed distinction between 1b and 1c will be very hard to 
make in an unambiguous way. Partly this may be due to the terminology, 
which is different from what I am used to from linguistics, where 
"morphosyntactic" properties are expressed by "inflectional" morphology, 
and "synthesis" is a technique (a way of assemling words) which in 
principle is neutral wrt the distinction between inflectional and 
derivational/lexical morphology. Since in the general case the boundary 
between inflectional and derivational/lexical morphology is far from 
clear, it will probably be more effort than it's worth to keep 
categories 1b and 1c separate. Individual languages (if they have a 
codified grammatical tradition, which most languages don't) will often 
have determined on more or leass clear grounds which morphological 
categories belong under which label (inflection or derivation)l, but 
cross-linguistically there are at the most tendencies, so that, e.g., 
number if expressed morphologically will tend to be an inflectional 
category.

The inclusion of inflectional and derivational morphology in the list 
also raises the vexed question of the nature of grammatical meaning. The 
use of "variant" for the number or case or other inflectional 
distinction implies that "in my houses" (as in Finnish inflected form 
_taloissani_ 'house plural inessive 1st person singular possessive') 
means more or less the same thing as "house", or am I reading to much 
into this? How about derivations like, e.g., "doer", "detainee", 
"eatery"? In other words: When you say "preserved semantic properties", 
is this also taken to imply "no added semantic properties"?

Best
Lars Borin


2012-08-20 18:34, John McCrae skrev:
> Hi all,
>
> Can I suggest we merge the following requirements on "Lexical Variant 
> and Paraphrases" and "Lexical and linguistic properties of lexical 
> entries"?
>
> My reasoning is that it seems that what Lupe is suggesting relies 
> heavily on the definition of properties. i.e., to model geographical 
> variants, register variants or diachronic variants, we need to be able 
> to state the geographical, register or diachronic properties of the 
> two variants. As such _we can think of variation in terms of the 
> properties that vary_ and those that do not. Put more clearly, 
> variants are entries that are similar (have the same property values) 
> except for some property, e.g., translation is variation in language, 
> pluralization is variation in number, etc.
>
> Considering the list of variants above, the following properties are 
> preserved by the type of variance
>
>   * Orthographic variants. Preserved: Pronounciation, syntax, most
>     pragmatic, semantic properties. Differs: Generally context or
>     geographic usage
>   * Inflectional variants. Preserved: part-of-speech, pragmatic and
>     semantic properties.
>   * Morphosyntactic variants. Preserved: semantic properties, most
>     pragmatic properties.
>   * Stylistic+Register variants. Preserved: semantics.
>   * Diachronic variants. Preserved: semantics
>   * Dialectical variants. Preserved: semantics
>   * Explicative variants. Preserved: extensional semantics (not
>     intensional)
>   * Semantic variants. Preserved: partial semantics
>
> As such I would go for splitting up the categories as follows
>
> *Group 1a. Orthographic variants*
>
>   * Historical Orthographic variants. e.g., different scripts such as
>     for Azeri (http://en.wikipedia.org/wiki/Azerbaijani_alphabet)
>   * Geo-orthographic variants. e.g., "localize" vs. "localise"
>   * Semantic-orthographics variants. e.g., "取る" (toru - "to take
>     (remove from a location)") vs "撮る" (toru - "to take (a photo)")
>
> *Group 1b. Inflectional variants*
>
>   * Pluralization, verb form inflection, comparatives and superlatives
>   * Synthesis (see http://en.wikipedia.org/wiki/Synthetic_language)
>
> *Group 1c. Morphosyntactic Variants*
>
>   * Rephrasing: e.g., "cancer of the mouth" vs. "mouth cancer"
>   * Derivation (e.g., Nominalization): e.g., "lexicon", "lexical",
>     "lexicalize"
>   * Pleonasm: "tuna" vs "tuna fish"
>   * Abbreviation: e.g., AIDS..... Philipp> Any variation has some
>     (sight) pragmatic implication, abbreviation for me is
>     morphosyntactic as the motivation is brevity rather than connotation.
>
> *Group 2a. Pragmatic Variants*
>
>   * stylistic or connotative variants (man and bloke)
>   * diachronic variants (tuberculosis and phthisis)
>   * dialectal variants (gasoline vs. petrol)
>   * pragmatic or register variants (headache and cephalalgia; swine
>     flu and pig flu and H1N1 and Mexic pandemic flu)
>
> *Group 2b. Circumlocutive variants*
>
>   * explicative variants (immigration law and law for regulating and
>     controlling immigration)
>
> *Group 3. Non-synonymous variants*
>
>   * Modification: "MRSA", vs "hospital-acquired MRSA"
>   * Hypernym/Hyponymy/Antonymy
>   * Cross-lingual narrowing/broadening: "river" vs "rivière/fleuve"
>
>
> Does this sound sensible or did I miss something?
>
> Regards,
> John
>
> On Sun, Aug 5, 2012 at 9:56 PM, lupe aguado <gac280771@gmail.com 
> <mailto:gac280771@gmail.com>> wrote:
>
>     Dear Ontolex members
>
>     With this message we would like to start the discussion about the
>     requirements on “Relations between lexical entries”. I put the
>     message as a draft in the Ontolex community Group and forgot to
>     send it to you. Sorry!
>
>     In our opinion, two types of relations need to be taken into
>     account in an ontology-lexicon model:
>
>      1. *relations between labels in different natural languages,* and
>      2. *relations between labels within the same natural language.*
>
>     Before continuing, we would like to define the two scenarios that
>     we envisage:
>
>      1. *A. **Multilingual labeling approach*
>
>     In a multilingual labeling approach, we have a single conceptual
>     structure, and we provide alternative labeling information in the
>     ontology-lexicon model for each of the languages covered (in the
>     same language or in different languages). This is possible
>     whenever the languages covered share a single view on a certain
>     domain. In this case, there will always be one or several labels
>     in each natural language for naming or terming the concepts in the
>     ontology.
>
>      1. *B. **Cross-lingual linking or mapping approach*
>
>     In this second scenario, there exist two independent monolingual
>     ontologies, defined in different languages, but covering the same
>     or similar subject domain. We aim at establishing links between
>     the labels that describe the two ontologies. The establishment of
>     these cross-lingual links could derive in cross-lingual ontology
>     mappings. In this scenario, the conceptual structure of each
>     ontology is modeled independently, and “linguistic links” or
>     “mappings” can be established between the two.
>
>     ---------
>
>     Now, in a *multilingual labeling approach*, we will usually refer
>     to “cross-lingual equivalents”. Let us take for example an
>     ontology of medical conditions. In such an ontology we can find
>     terms such as menopause in English, and its cross-lingual
>     equivalents: menopause in French, menopause in Danish,
>     vaihdevuodet in Finnish or Menopause in German. This means that
>     the “same” concept exists in the involved cultures and has an
>     equivalent term in the corresponding language.
>
>     On the contrary, in a *cross-lingual linking or mapping approach*,
>     we could come across several types of relations among lexical
>     entries due to the following reasons:
>
>       * conceptualization mismatches
>       * different levels of granularity
>
>     In fact, granularity or viewpoint differences may also come up in
>     a “monolingual” linking or mapping approach. However,
>     conceptualization mismatches will be more common in a
>     cross-lingual scenario. In this sense, we could account for
>     several types of relations
>
>     1. *Cross-lingual equivalence relations*, as in the multilingual
>     labeling scenario. These would establish a relation between
>     concepts that are not exactly the same (do not have the same
>     intension and/or extension), but are close equivalents, because no
>     exact equivalent exists. Example: full professor in English –
>     catedrático in Spanish – Professor in German. In order to
>     distinguish them from the cross-lingual equivalents in the
>     multilingual labeling scenario, we could term them: *cross-lingual
>     close equivalents*? *Cross-lingual near equivalents*? Suggestions
>     are welcome!!
>
>     2. *Cross-lingual broad (narrow) equivalence relations*. These
>     would establish a relation between concepts with different levels
>     of granularity. This usually happens when one culture understands
>     a concept or phenomenon with a higher granularity than the other,
>     i.e., one culture has two or more concepts (and in its turn, terms
>     for naming them) to describe the same phenomenon. Example: river
>     in English – rivière and fleuve in French; Tötung in German –
>     asesinato and homicidio in Spanish. Here again, suggestions for
>     better examples are welcome.
>
>     In the case no equivalent exists, we could still provide a term or
>     description, using for this a mixed scenario, i.e., providing some
>     labels or lexical entries for the concept we do not find an
>     equivalent term in the other ontology, as in the multilingual
>     labeling approach. For this, we consider two options:
>
>     3. *Literal translation relations*. These are translations of
>     terms that describe concepts that do not exist in the target
>     language, and for which a literal or “word for word translation”
>     is provided so that the concept is understood by the target
>     language. Example: École normal in French– (French) Normal School
>     in English; Presidente del Gobierno in Spanish – President of the
>     Government in English.
>
>     4. *Descriptive translation relations.* These are translations of
>     terms that describe concepts that do not exist in the target
>     language, and for which a description or definition (and not a
>     term) is provided in the target language. Example: Panetone in
>     Italian – bizcocho italiano que se consume en Nochevieja in
>     Spanish. In this case, we could also opt for repeating the Italian
>     Word plus the gloss.
>
>     In the latter two cases, we could also provide a link to the
>     closest equivalent or superclass (by means of the cross-lingual
>     broad equivalence relation), and additionally provide a literal or
>     descriptive translation.
>
>     -------
>
>     As for the *relations* *between labels within the same language*,
>     we propose to talk about “term variation”.  For example:  what is
>     the difference between Advertising and Publicity, if any? And
>     between Contamination and Pollution?, or between Assisted
>     conception, Artificial insemination and in vitro Fertilization? In
>     a SKOS Thesaurus, Assisted conception is the main label, and the
>     rest are alternative labels. However, we think that we could be
>     more specific regarding the type of variants pointing to one and
>     the same concept in the ontology, and that this should be
>     accounted for in our ontology-lexicon model. Sometimes, the
>     difference is a consequence of the contextual (pragmatic) usage,
>     and we have to decide whether to represent this in our model.
>
>     Based on previous classifications of terminology variation, we
>     have identified three main groups of term variants that include
>     the following types (see also [1] and [2]):
>
>     *Group 1*. Synomyms or terminological units that totally
>     correspond to the same concept:
>
>       * graphical and orthographical variants (/localization
>         /and/localisation/);
>       * inflectional variants (/cat/ and /cats/);
>       * morphosyntactic variants (/nitrogen fixation/ and /fixation of
>         nitrogen/).
>
>     *Group 2*. Partial synonyms or terminological units that highlight
>     different aspects of the same concept:
>
>       * stylistic or connotative variants (/man/ and /bloke/)
>       * diachronic variants (/tuberculosis/ and /phthisis/)
>       * dialectal variants (/gasoline/ vs. /petrol/)
>       * pragmatic or register variants (/headache/ and /cephalalgia/;
>         /swine flu/ and /pig flu/ and /H1N1/ and /Mexic pandemic flu/)
>       * explicative variants (/immigration law/ and /law for
>         regulating and controlling immigration/)
>
>     So, we would be very grateful for your suggestions and comments on
>     this proposal.
>
>     Best regards,
>
>     Lupe and Elena
>
>     [1] Montiel-Ponsoda, E., Aguado de Cea, G., McCrae, J. (2011).
>     Representing term variation in /lemon/. In Proceedings of the /WS
>     2Ontology and lexicon: new insights, TIA 2011 - 9th International
>     Conference on Terminology and Artificial Intelligence/, pp. 47–50.
>
>     [2] Aguado de Cea, G., and Montiel-Ponsoda, E. (2012).  Term
>     variants in ontologies. In Proceedings of the AESLA (/Asociación
>     Española de Lingüística Aplicada/) Conference.
>
>
>
>
>     2012/7/18 Philipp Cimiano <cimiano@cit-ec.uni-bielefeld.de
>     <mailto:cimiano@cit-ec.uni-bielefeld.de>>
>
>         Dear all,
>
>          and just to clarify what the description of the requirements
>         should include:
>
>         Under "Description", there should be a general description of
>         the requirement, its implications, etc. It is important that
>         we think here in terms of requirements on the general model,
>         not on particular data categories, properties, etc. but on
>         requirements at the meta-model level.
>
>         Under "Relevant Use Cases": here we should just list the IDs
>         of the use cases touched by this requirement. Maybe this
>         should be called "Affected Use Cases" ???
>
>         "Relation to Use Case": here we should give detailed examples
>         from the use cases where the requirement is important, thus
>         grounding our requirements in the use cases we have collected.
>
>         If there are any questions on this, just shoot.
>
>         Best regards,
>
>         Philipp.
>
>
>
>         Am 18.07.12 14:24, schrieb Philipp Cimiano:
>
>             Dear ontolex members,
>
>              during our last meeting on the 6th of July, we discussed
>             my condensed list of requirements on the model and agreed
>             that it looks promising to work on the basis of these from
>             now on.
>
>             See here:
>             http://www.w3.org/community/ontolex/wiki/Specification_of_Requirements
>
>             The older list of unstructured requirements is linked from
>             the bottom of the page.
>
>             We fixed the following responsibles to produce a first
>             draft of the requirement and kick-off the discussions on
>             this mailinglist. (We really need to start the discussion
>             on the relevant issues!)
>
>             - Express Meaning with respect to ontology:
>             John/Philipp/Aldo/Guido
>             - Lexical Variation and Paraphrases: Philipp
>             - Relation between lexical entries: Lupe/Elena
>             - Lexical and linguistic properties of lexical entries:
>             John/Philipp
>             - Valence and Ontological Mapping: John/Philipp
>             - High-Order Predicate Mapping: John/Philipp
>             - Lexico-Syntactic Patterns: Elena/Dagmar
>             - Metadata about lexicon: Armando
>             - Modelling lexical resources: John/Aldo
>
>             The goal would be to have a detailed specification and an
>             ongoing discussion on this mailinglist by end of August.
>
>             The next teleconference will be on September 6th, 15:00 -
>             17:00 (CET). It will be two hours as we decided to skip
>             the one in August due to holiday period.
>
>             We also decided to have biweekly teleconferences from
>             September on. I think it is important to keep things
>             moving quickly. Otherwise I have the feeling that not much
>             happens in between our monthly teleconferences.
>
>             I am now on holidays for two weeks and will then start
>             working on the requirements assigned to me.
>             Needless to say, everyone should feel free to start
>             working on their requirements as soon as possible.
>
>             If you think that an important requirement is missing,
>             please post it on the list and we will discuss it.
>
>             Best regards,
>
>             Philipp.
>
>
>
>         -- 
>         Prof. Dr. Philipp Cimiano
>         Semantic Computing Group
>         Excellence Cluster - Cognitive Interaction Technology (CITEC)
>         University of Bielefeld
>
>         Phone: +49 521 106 12249 <tel:%2B49%20521%20106%2012249>
>         Fax: +49 521 106 12412 <tel:%2B49%20521%20106%2012412>
>         Mail: cimiano@cit-ec.uni-bielefeld.de
>         <mailto:cimiano@cit-ec.uni-bielefeld.de>
>
>         Room H-127
>         Morgenbreede 39
>         33615 Bielefeld
>
>
>
>

-- 
«Null hull,» sa Harry    | – Bögga? sagði Erlendur. Er það orð?
(Jo Nesbø: Kakerlakkene) | (Arnaldur Indriðason: Mýrin)
--
Lars Borin
Språkbanken • Centre for Language Technology
Institutionen för svenska språket
Göteborgs universitet
Box 200
SE-405 30 Göteborg
Sweden

office +46 (0)31 786 4544
mobile +46 (0)70 747 8386

<http://språkbanken.gu.se/personal/lars/>