W3C home > Mailing lists > Public > public-esw-thes@w3.org > February 2010

Re: Unicode Characters to display SKOS relations

From: Stella Dextre Clarke <stella@lukehouse.org>
Date: Mon, 08 Feb 2010 12:56:51 +0000
Message-ID: <4B700A13.6040500@lukehouse.org>
To: Christophe Dupriez <christophe.dupriez@destin.be>, SKOS <public-esw-thes@w3.org>
CC: Simon Spero <ses@unc.edu>, Ed Summers <ehs@pobox.com>, Seth van Hooland <svhoolan@ulb.ac.be>, Bhojaraju Gunjal <bhojaraju.g@gmail.com>, Emmanuel Di Pretoro <edipretoro@gmail.com>, José Ramón Pérez Agüera <jose.aguera@gmail.com>, danbrickley@gmail.com
Christophe's ideas for icons look very promising. You may be interested 
to compare with ideas-in-progress in ISO NP 25964.

We are currently debating and drafting ISO 25964 Part 2, which deals 
with mapping between thesauri, and between thesauri and other sorts of 
vocabulary. We are hoping to provide for mapping types which align 
reasonably well with those in SKOS (although we hope also to provide for 
one-to-many mappings, not yet represented in SKOS).

We come from the long-standing tradition of representing relationships 
within one thesaurus with the following tags:

USE/UF		equivalence between terms
BT		broader term (really broader concept)
NT		narrower term (really narrower concept)
RT		related term (really related concept)

If you can imagine a mixed display to include these relationships 
alongside mappings between thesauri, it seems useful to distinguish them 
in some may, so that internal relationships are not confused with 
external mappings. Therefore we are considering:

EQ		equivalence mappings (between concepts)
BM		broader mapping  (or broader match)
NM		narrower mapping (or narrower match)
RM		related mapping (or related match)

So far, so simple; but what about inexact equivalence? We are now 
considering introduction of the tilde (~) to indicate inexactitude. Thus 
"EQ~" or "~EQ" would show an inexact equivalence mapping. The definition 
of this type of mapping is not far from the SKOS notion of closeMatch. 
And there is a reasonable analogy in Maths for this symbol.

I'm mentioning this in response to Christophe's proposals (which in 
general seem a good idea), because our tilde would not align well with 
the way he proposes to use rather similar (but more elaborate) symbols, 
and that could slightly upset our hopes for alignment with SKOS. I don't 
think we'll venture into the amazing range of icons he has proposed, but 
it would be nice to avoid clashes.

I should stress that our ideas for tags/symbols, and indeed for the 
mappings themselves, are by no means final. This is a good stage for 
comments from anyone on any aspect of the above proposals, because we 
still have room for manoeuvre. That said, I'd like to keep the tilde if 
we can - short, sweet and simple, and it's on our keyboards.

Any comment?
Stella Dextre Clarke,
Project Leader, ISO NP 25964
*****************************************************
Stella Dextre Clarke
Information Consultant
Luke House, West Hendred, Wantage, OX12 8RR, UK
Tel: 01235-833-298
Fax: 01235-863-298
stella@lukehouse.org
*****************************************************





Christophe Dupriez wrote:
> Dear Simon,
> 
> You are right: my goal is to make big displays of thesauri legible, not 
> to invent new mathematics.
> Using well known mathematical symbols are problematic. Unicode implies 
> reuse of existing symbols, already loaded with some meanings.
> 
> So I updated my proposal again to take your remark. Please (re)look at:
> http://www.destin.be/ASKOSI/Wiki.jsp?page=Icons%20for%20SKOS
> 
> I did not finalize icons for "matching" relations but I will following 
> the next wave of comments.
> 
> New icons:
> Concept:
> ConceptScheme:
> Broader:
> Narrower:
> Related:
> 
> Example:
> http://www.destin.be/ASKOSI/Wiki.jsp?page=Icons%20for%20SKOS#section-Icons+for+SKOS-WithIcons
> 
> Thanks for the references (especially Dagobert Soergel which works like 
> me for Digital Libraries).
> 
> Have a nice w.e.
> 
> Christophe
> 
> Simon Spero a écrit :
>> ----
>> Useful sources:
>>
>> Croft, William and D. Alan Cruse (2004). /Cognitive linguistics/. 
>> Cambridge University Press.
>>
>> Cruse, D. Alan. (1986). /Lexical semantics/. Cambridge University Press.
>>
>> Riesthuis, Gerhard J. A. et al. (2008). /Guidelines for Multilingual 
>> Thesauri/. IFLA Professional Reports,
>> No. 115. The Hague, NL: International Federation of Library 
>> Associations and Institutions. URL:
>> http://www.ifla.org/VII/s29/pubs/Profrep115.pdf
>>
>> Soergel, Dagobert (1974). Indexing languages and thesauri: 
>> construction and maintenance. Information sciences series. Los 
>> Angeles: Melville Pub. Co. ISBN: 0471810479.
>>
>> In fact, read as much Soergel as you can find :-)
>>
>> ----------
>>
>> I'm not sure if the semantics of SKOS are quite right for the 
>> mathematical symbols  you're using
>>
>> Let the type Δ denote the domain of discourse to which labels might be 
>> attached.
>> Let the type Σ denote the set of all possible label strings .
>> Let the type CONCEPT denote a two-tuple, (Σ × ℙ(Δ))   containing a 
>> label and a set of elements of Δ .
>> Let the type CONCEPT-SCHEME denote a  2-tuple (Σ  ×  ℙ(CONCEPT)).
>>
>> Let c and k denote two arbitrary CONCEPTs
>> Let C and K denote two arbitrary CONCEPT-SCHEMEs
>>
>> Let label(k) refer to the first element of CONCEPT k.
>> Let documents(k) refer to the second element of CONCEPT k.
>> Let label(C) refer to the first element of a CONCEPT-SCHEME C.
>> Let concepts(C) refer to the second element of a CONCEPT-SCHEME C.
>>
>> Let the 2-tuple (c,C) denote a fully qualified concept (FQC) 
>> consisting of  of a concept and a concept scheme,  where c ∈ concepts(C).
>>
>> BT, NT, and EQ for a single CONCEPT scheme.
>>
>> Within a single CONCEPT-SCHEME C, such that c ∈ C ⋀ k ∈ C
>>
>> 1: ( BT)    c < k    iff documents(c) ⊂ documents(k)
>> 2: (NT)    c > k     iff documents(c) ⊃ documents(k)
>> 3: (SY)    c  ≍ k    iff documents(c) ≡ documents(k)
>>
>> Unique Concept Scheme Name Assumption
>> 4: ∀ C ∈ CONCEPT-SCHEME. ∀ K ∈ CONCEPT-SCHEME. label(C) ≡ label(D)  → 
>> C ≡ D
>>
>> Within Scheme Unique Preferred Name Assumption
>> 5: ∀ C ∈ CONCEPT-SCHEME. ∀ c ∈ concepts(C).  ∀ d ∈ concepts(C).  
>> label(c) = label(d) → c ≡ d
>>
>> Identity
>> 6: C = K    iff   label(c) ≡ label(k) ⋀  concepts(c) ≡ concepts(k)
>> 7: c = k     iff   label(c) ≡ label(k) ⋀  documents(c) ≡ documents(k) 
>> ^ ∀ C ∈ CONCEPT-SCHEME. c ∈ concepts(C) iff k ∈ concepts(C)
>>
>>
>> Mapping Relations
>>    Note that mapping relations are only defined between concepts in 
>> different concept schemes.
>>
>> Exact match, (c,C) ≍ (k,K)
>> 8: For an exact match, (c,C) ≍ (k,K)
>>     (c,C) ≍ (k,K) iff C ≢ K ⋀  documents(c) ≡ documents(k)
>>
>> Broad Match:  (c,C)  ⪷ (k,K)
>>
>> 9: (c,C) ⪷ (k,K)  iff
>>               ¬ (c,C) ≍ (k,K) ⋀
>>                c < k  ⋀
>>               ∄d ∈ concepts(K). (c < d ⋀ d < k  ⋀ d ≠ k)
>>
>> Narrower Match: (c,C) ⪸ (k,K)
>>
>> 10: (c,C) ⪸ (k,K)  iff
>>               ¬ (c,C) ≍ (k,K) ⋀
>>                c > k  ⋀
>>               ∄d ∈ concepts(K). (c > d ⋀ d > k  ⋀ d ≠ k)
>>
>>
>> Close Match:  (c,C) ≈ (k,K)
>>
>> The semantics of close match are under determined:  as a bare minimum, 
>> we must define a similarity function  f ∈ (CONCEPT × CONCEPT → [0,1]), 
>> together with a threshold t below which two concepts are not 
>> considered to be a match.
>>
>> 11:    (c,C) ≈ (k,K) iff  ¬ (c,C) ≍ (k,K) ⋀
>>                               f(c,k) ≥ t   ⋀
>>                               ∄d ∈ concepts(K). f(c,d) > f(c,k)
>>
> 


-- 
*****************************************************
Stella Dextre Clarke
Information Consultant
Luke House, West Hendred, Wantage, OX12 8RR, UK
Tel: 01235-833-298
Fax: 01235-863-298
stella@lukehouse.org
*****************************************************
Received on Monday, 8 February 2010 12:57:28 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 2 March 2016 13:32:13 UTC