IVOA

Vocabularies in the Virtual Observatory
Version wd-1.0

IVOA Working Draft, 2008 May 8 [Editors' draft, Revision: 420 ]

This version
http://www.ivoa.net/Documents/WD/Semantics/vocabularies-2008-05-08.html
Latest version
http://www.ivoa.net/Documents/latest/vocabularies.html
and issues list
Previous version
http://www.ivoa.net/Documents/WD/Semantics/vocabularies-20080320.html
Working Group
Semantics
Editors
Alasdair J G Gray, University of Glasgow, UK
Norman Gray, University of Leicester / University of Glasgow, UK
Frederic V Hessman, University of Göttingen, Germany
Andrea Preite Martinez, INAF, Italy
Authors
Sébastien Derriere, Alasdair J G Gray, Norman Gray, Frederic V Hessman, Tony Linde, Andrea Preite Martinez, Rob Seaman and Brian Thomas

Abstract [link here]

As the astronomical information processed within the Virtual Observatory becomes more complex, there is an increasing need for a more formal means of identifying quantities, concepts, and processes not confined to things easily placed in a FITS image, or expressed in a catalogue or a table. We proposed that the IVOA adopt a standard format for vocabularies based on the W3C's Resource Description Framework (RDF) and Simple Knowledge Organisation System (SKOS). By adopting a standard and simple format, the IVOA will permit different groups to create and maintain their own specialised vocabularies while letting the rest of the astronomical community access, use, and combine them. The use of current, open standards ensures that VO applications will be able to tap into resources of the growing semantic web. Several examples of useful astronomical vocabularies are provided, including work on a common IVOA thesaurus intended to provide a semantic common base for VO applications.

Status of this document [link here]

This is an IVOA Working Draft. The first release of this document was 2008 March 20.

This document is an IVOA Working Draft for review by IVOA members and other interested parties. It is a draft document and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use IVOA Working Drafts as reference materials or to cite them as other than “work in progress”.

A list of current IVOA Recommendations and other technical documents can be found at http://www.ivoa.net/Documents/.

Acknowledgments

We would like to thank the members of the IVOA semantic working group for many interesting ideas and fruitful discussions.

Table of Contents

1. Introduction (informative) [link here]

1.1. Vocabularies in astronomy [link here]

Astronomical information of relevance to the Virtual Observatory (VO) is not confined to quantities easily expressed in a catalogue or a table. Fairly simple things such as position on the sky, brightness in some units, times measured in some frame, redshifts, classifications or other similar quantities are easily manipulated and stored in VOTables and can currently be identified using IVOA Unified Content Descriptors (UCDs) [std:ucd]. However, astrophysical concepts and quantities use a wide variety of names, identifications, classifications and associations, most of which cannot be described or labelled via UCDs.

There are a number of basic forms of organised semantic knowledge of potential use to the VO, ranging from informal “folksonomies” (where users are free to choose their own labels) at one extreme, to formally structured “vocabularies” (where the label is drawn from a predefined set of definitions, and which can include relationships between labels) and “ontologies” (where the domain is captured in a formal data model) at the other. More formal definitions are presented later in this document.

An astronomical ontology is necessary if we are to have a computer (appear to) “understand” something of the domain. There has been some progress towards creating an ontology of astronomical object types [std:ivoa-astro-onto] to meet this need. However there are distinct use cases for letting human users find resources of interest through search and navigation of the information space. The most appropriate technology to meet these use cases derives from the Information Science community, that of controlled vocabularies, taxonomies and thesauri. In the present document, we do not distinguish between controlled vocabularies, taxonomies and thesauri, and use the term vocabulary to represent all three.

One of the best examples of the need for a simple vocabulary within the VO is VOEvent [std:voevent], the VO standard for supporting rapid notification of astronomical events. This standard requires some formalised indication of what a published event is “about”, in a formalism which can be used straightforwardly by the developer of relevant services. See 1.2. Use-cases, and the motivation for formalised vocabularies for further discussion.

A number of astronomical vocabularies have been created, with a variety of goals and intended uses. Some examples are detailed below.

1.2. Use-cases, and the motivation for formalised vocabularies [link here]

The most immediate high-level motivation for this work is the requirement of the VOEvent standard [std:voevent] for a controlled vocabulary usable in the VOEvent's <Why/> and <What/> elements, which describe what sort of object the VOEvent packet is describing, in some broadly intelligible way. For example a “burst” might be a gamma-ray burst due to the collapse of a star in a distant galaxy, a solar flare, or the brightening of a stellar or AGN accretion disk, and having an explicit list of vocabulary terms can help guide the event publisher into using a term which will be usefully precise for the event's consumers. A free-text label can help here (which brings us into the domain sometimes referred to as folksonomies), but the astronomical community, with a culture sympathetic to international agreement, can do better.

The purpose of this proposal is to establish a set of conventions for the creation, publication, use, and manipulation of astronomical vocabularies within the Virtual Observatory, based upon the W3C's SKOS standard. We include as appendices to this proposal formalised versions of a number of existing vocabularies, encoded as SKOS vocabularies [std:skosref].

Specific use-cases include the following.

The goal of this standard is to show how vocabularies can be easily expressed in an interoperable and computer-manipulable format, and the sole normative section of this Recommendation (namely section 3. Publishing vocabularies (normative)) contains requirements and suggestions intended to promote this. Four example vocabularies that have previously been expressed using non-standardized formats – namely the A&A keyword list, the IAU and AOIM thesauri, and UCD1 – are included below as illustrations of how simple it is to publish them in SKOS, without losing any of the information of the original source vocabularies.

It is not a goal of this standard, as it is not a goal of SKOS, to produce knowledge-engineering artefacts which can support elaborate machine reasoning – such artefacts would be very valuable, but require much more expensive work on ontologies. As the supernova use-case above illustrates, even simple vocabularies can support useful machine reasoning.

It is also not a goal of this standard to produce new vocabularies, or substantially alter existing ones; instead, the vocabularies included below in section 4. Example vocabularies (informative) are directly derived from existing vocabularies (the exceptions are the IVOAT vocabulary, which is ultimately intended to be a significant update to the IAU-93 original, and the constellations vocabulary, which is intended to be purely didactic). It therefore follows that the ambiguities, redundancies and incompleteness of the source vocabularies are faithfully represented in the distributed SKOS vocabularies. We hope that this formalisation process will create greater visibility and broader use for the various vocabularies, and that this will guide the maintenance efforts of the curating groups.

The reason for both of these limitations is that vocabularies are extremely expensive to produce, maintain and deploy, and we must therefore rely on such vocabularies as have been developed, and attached as metadata to resources, by others. Such vocabularies are less rich or less coherent than we might prefer, but they are widely enough deployed to be useful. We hope that the set of example vocabularies we have provided will build on this deployment, by providing material which is useful out of the box.

1.3. Formalising and managing multiple vocabularies [link here]

We find ourselves in the situation where there are multiple vocabularies in use, describing a broad range of resources of interest to professional and amateur astronomers, and members of the public. These different vocabularies use different terms and different relationships to support the different constituencies they cater for. For example, “delta Sct” and “RR Lyr” are terms one would find in a vocabulary aimed at professional astronomers, associated with the notion of “variable star”; however one would not find such technical terms in a vocabulary intended to support outreach activities.

One approach to this problem is to create a single consensus vocabulary, which draws terms from the various existing vocabularies to create a new vocabulary which is able to express anything its users might desire. The problem with this is that such an effort would be very expensive, both in terms of time and effort on the part of those creating it, and to the potential users, who have to learn to navigate around it, recognise the new terms, and who have to be supported in using the new terms correctly (or, more often, incorrectly).

The alternative approach to the problem is to evade it, and this is the approach taken in this document. Rather than deprecating the existence of multiple overlapping vocabularies, we embrace it, help interest groups formalise as many of them as are appropriate, and standardise the process of formally declaring the relationships between them. This means that:

2. SKOS-based vocabularies (informative) [link here]

In this section, we introduce the concepts of SKOS-based vocabularies, and the technology of mapping between them. We describe some additional requirements for IVOA vocabularies in the next section, 3. Publishing vocabularies (normative).

2.1. Selection of the vocabulary format [link here]

After extensive online and face-to-face discussions, the authors have brokered a consensus within the IVOA community that formalised vocabularies should be published at least in SKOS (Simple Knowledge Organisation System) format, a W3C draft standard application of RDF to the field of knowledge organisation [std:skosref]. SKOS draws on long experience within the Library and Information Science community, to address a well-defined set of problems to do with the indexing and retrieval of information and resources; as such, it is a close match to the problem this document is addressing.

ISO 5964 [std:iso5964] defines a number of the relevant terms (ISO 5964:1985=BS 6723:1985; see also [std:bs8723-1] and [std:z39.19]), and some of the (lightweight) theoretical background. The only technical distinction relevant to this document is that between vocabulary and thesaurus: BS-8723-1 defines a thesaurus as a

Controlled vocabulary in which concepts are represented by preferred terms, formally organized so that paradigmatic relationships between the concepts are made explicit, and the preferred terms are accompanied by lead-in entries for synonyms or quasi-synonyms. (BS-8723-1, sect. 2.39)

with a similar definition in ISO-5964 sect. 3.16. The paradigmatic relationships in question are those relating a term to a “broader”, “narrower” or more generically “related” term. These notions have an operational definition: any resource retrieved as a result of a search on a given term will also be retrievable through a search on that term's “broader term” (“narrower” is a simple inverse, so that for any pair of terms, if A skos:broader B, then B skos:narrower A; a term may have multiple narrower and broader terms). This is not a subsumption relationship, as there is no implication that the concept referred to by a narrower term is of the same type as a broader term.

Thus a vocabulary (SKOS or otherwise) is not an ontology. It has lighter and looser semantics than an ontology, and is specialised for the restricted case of resource retrieval. Those interested in ontological analyses can easily transfer the vocabulary relationship information from SKOS to a formal ontological format such as OWL [std:owl].

The purpose of a thesaurus is to help users find resources they might be interested in, be they library books, image archives, or VOEvent packets.

2.2. Content and format of a SKOS vocabulary [link here]

A published vocabulary in SKOS format consists of a set of “concepts” – an example concept capturing the vocabulary information about spiral galaxies is provided in the Figure below, with the RDF shown in both RDF/XML [std:rdfxml] and Turtle notation [std:turtle] (Turtle is similar to the more informal Notation3). The elements of a concept are detailed below.

Figure: examples of SKOS vocabularies

XML Syntax Turtle Syntax 
<skos:Concept rdf:about="#spiralGalaxy">
  <skos:prefLabel lang="en">
    spiral galaxy
  </prefLabel>
  <skos:prefLabel lang="de">
    Spiralgalaxie
  </prefLabel>
  <skos:altLabel lang="en">
    spiral nebula
  </skos:altLabel>
  <skos:hiddenLabel lang="en">
    spiral glaxy
  </hiddenLabel>
  <skos:definition lang="en">
    A galaxy having a spiral structure.
  </skos:definition>
  <skos:scopeNote lang="en">
    Spiral galaxies fall into one of 
    three catagories: Sa, Sc, and Sd.
  </skos:scopeNote>
  <skos:narrower
    rdf:resource="#barredSpiralGalaxy"/>
  <skos:broader
    rdf:resource="#galaxy"/>
  <skos:related
    rdf:resource="#spiralArm"/>
</skos:Concept>

<#spiralGalaxy> a skos:Concept;
  skos:prefLabel
    "spiral galaxy"@en, 
    "Spiralgalaxie"@de;
  skos:altLabel "spiral nebula"@en;
  skos:hiddenLabel "spiral glaxy"@en;
  skos:definition """A galaxy having a 
    spiral structure."""@en;
  skos:scopeNote """Spiral galaxies fall
    into one of three categories:
    Sa, Sc, and Sd"""@en;
  skos:narrower <#barredSpiralGalaxy>;
  skos:broader <#galaxy>;
  skos:related <#spiralArm> .

A SKOS vocabulary includes the following features.

In addition to the information about a single concept, a vocabulary can contain information to help users navigate its structure and contents:

2.3. Mapping relationships between vocabularies [link here]

There already exist several vocabularies in the domain of astronomy. Instead of attempting to replace all these existing vocabularies, which have been developed to achieve different aims and user groups, we embrace them. This requires a mechanism to relate the concepts in the different vocabularies.

Part of the SKOS standard [std:skosref] allows a concept in one vocabulary to be related to a concept in another vocabulary. There are four types of relationship provided to capture the relationships between concepts in vocabularies, which are similar to those defined for relationships between concepts within a single vocabulary. The types of mapping relationships are:

The semantic mapping relationships have certain properties. The broadMatch relationship has the narrowMatch relationship as its inverse and the exactMatch and relatedMatch relationships are symmetrical. The consequence of these properties is that if you have a mapping from concept A in one vocabulary to concept B in another vocabulary then you can infer a mapping from concept B to concept A.

At the time of writing, the SKOS document is still a draft, and may or may not end up with support for mappings in the core document rather than in a companion document. This section of this Working Draft, and other references to mappings below, should therefore be considered provisional until it becomes clear how best to implement the eventual SKOS guidelines.

2.4. Vocabulary versions [link here]

The document [kendall08] discusses good practice for managing RDF vocabularies. At the time of writing (2008 May) this is still an editor's draft, and it itself notes that good practice in this area is not yet fully stable, so our recommendations here are necessarily tentative, and in some places restricted to the relatively small vocabularies (100s to 1000s of terms) we expect to encounter in the VO. We expect to adjust or enhance this advice in future editions of this Recommendation, as best practice evolves, or as we gain more experience with the relevant vocabularies.

We must distinguish between versions of a vocabulary, and versions of the description of a vocabulary. In the former case, we are concerned with the presence or absence of certain concepts, such as “star” or “GRB”, and expect that there will be some reasonably stable relationship between the concept URI and the real-world concept it refers to. In the latter case, we are concerned with the technicalities of associating a concept URI with its labels, its description, and with other related concept URIs. While it is true that there are epistemological commitments involved in the simple act of naming (and the terms “GRB” and “planet” remind us that there is knowledge implicit within a name), it is the latter case that generally represents the knowledge we have of an object, and it is this knowledge which we must version.

In consequence, the concept URIs should not carry version information. The partial exception to this is when a vocabulary undergoes a major restructuring, as a result of the terms in it becoming significantly incoherent – for example, we might imagine the IAU92 thesaurus being updated to form an IAU 200x thesaurus – but in this case we should regard the result as a new vocabulary, rather than simply an adjusted version of an old one.

The SKOS vocabulary has all of its terms appearing in the same unversioned base namespace (the year and month in that namespace is a consequence of the management of the W3C URI namespace, and is not connected to SKOS versions), and once there is not removed [kendall08] (there seems to be no discussion of this in a SKOS document, as opposed to commentary on SKOS). Successive versions of the vocabulary description describe the vocabulary terms as “unstable”, “testing”, “stable” or “deprecated”.

The Dublin Core namespaces are managed in a similar way [dc:namespaces]. The namespace URIs, which act as common prefixes to the DC terms, and which are defined using a “hash URI” strategy, in RDF terms, have no version numbers, so that the namespace for the DC terms vocabulary is http://purl.org/dc/terms/. Terms such as http://purl.org/dc/terms/extent then 302-redirect to a URL which, for administrative convenience, happens to contain a release date, but which resolves to RDF which defines the unversioned term http://purl.org/dc/terms/extent. This file includes the following content (translated into Turtle from the original RDF/XML for legibility).

@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix skos: <http://www.w3.org/2004/02/skos/core#> .
@prefix dcam: <http://purl.org/dc/dcam/> .
@prefix dcterms: <http://purl.org/dc/terms/> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .

<http://purl.org/dc/terms/>
  dcterms:title """DCMI Namespace for metadata terms
                in the http://purl.org/dc/terms/ namespace"""@en-us;
  rdfs:comment """To comment on this schema, 
               please contact dcmifb@dublincore.org.""";
  dcterms:publisher "The Dublin Core Metadata Initiative"@en-us;
  dcterms:modified "2008-01-14" .

dcterms:extent
  rdfs:label "Extent"@en-us;
  rdfs:comment "The size or duration of the resource."@en-us;
  rdfs:isDefinedBy <http://purl.org/dc/terms/>;
  dcterms:issued "2000-07-11";
  dcterms:modified "2008-01-14";
  a rdf:Property;
  dcterms:hasVersion <http://dublincore.org/usage/terms/history/#extent-003>;
  rdfs:range dcterms:SizeOrDuration;
  rdfs:subPropertyOf <http://purl.org/dc/elements/1.1/format>,
    dcterms:format .
...

This includes the definition of the (unversioned) http://purl.org/dc/terms/extent concept, along with semantic knowledge about the concept (rdfs:subPropertyOf) as of 2008-01-14, plus other editorial (dcterms:modified) and definitional (rdfs:isDefinedBy) metadata.

3. Publishing vocabularies (normative) [link here]

3.1. Requirements [link here]

A vocabulary which conforms to this IVOA standard has the following features. In this section, the keywords must, should and so on, are to be interpreted as described in [std:rfc2119].

3.1.1. Dereferenceable namespace

The namespace of the vocabulary must be dereferenceable on the web. That is, typing the namespace URL into a web browser will produce human-readable documentation about the vocabulary. In addition, the namespace URL should return an RDF version of the vocabulary if it is retrieved with one of the RDF MIME types in the HTTP Accept header. At the time of writing, the only fully standardised RDF MIME type is application/rdf+xml for RDF/XML, but text/rdf+n3 and text/turtle are the proposed types for Notation3 [notation3] and Turtle [std:turtle], respectively.

Rationale: These prescriptions are intended to be compatible with the patterns described in [berrueta08] and [sauermann08], and vocabulary distributors should follow these patterns where possible.

3.1.2. Long-term availability

The files defining a vocabulary, including those of superseded versions, should remain permanently available. There is no requirement that the namespace URL be at any particular location, although the IVOA web pages, or a journal publisher's web pages, would likely be suitable archival locations.

3.1.3. Distribution format

Vocabularies must be made available for distribution as SKOS RDF files, in either RDF/XML [std:rdfxml] or Turtle [std:turtle] format; vocabularies should be made available in both formats. As an alternative to Turtle, vocabularies may be made available in that subset of Notation3 [notation3] which is compatible with Turtle; if Turtle or Notation3 is being served, it is prudent to support both text/rdf+n3 and text/turtle as MIME types in the Accept header of the HTTP request. See issue [distformat-2].

A publisher may make available RDF in other formats, or other supporting files. A publisher must make available at least some human-readable documentation – see section 3.3. Good practices when serving vocabularies on the web for a discussion of the mechanics here.

Rationale: this does imply that the vocabulary source files can only realistically be parsed using an RDF parser. An alternative is to require that vocabularies be distributed using a subset of RDF/XML which can also be naively handled as traditional XML; however as well as creating an extra standardisation requirement, this would make it effectively infeasible to write out the distribution version of the vocabulary using an RDF or general SKOS tool.

3.1.4. Clearly versioned vocabulary [link here]

The vocabulary namespace should not be versioned, but it should be easy to retrieve earlier versions of the RDF describing the vocabulary. See the discussion in section 2.4. Vocabulary versions for the rationale for this, and see section 3.3. Good practices when serving vocabularies on the web for a discussion of its implications for the way that vocabularies are served on the web.

3.1.5. No restrictions on source files

This Recommendation does not place any restrictions on the format of the files managed by the maintenance process, as long as the distributed files are as specified above. See issue [masterformat-1].

3.2. Good practices of vocabulary design [link here]

This standard imposes a number of requirements on conformant vocabularies (see 3.1. Requirements). In this section we list a number of good practices that IVOA vocabularies should abide by. Some of the prescriptions below are more specific than good-practice guidelines for vocabularies in general.

The adoption of the following guidelines will make it easier to use vocabularies in generic VO applications. However, VO applications should be able to accept any vocabulary that complies with the latest SKOS standard [std:skosref] (this is a syntactical requirement, and does not imply that an application will necessarily understand the terms in an alien vocabulary, although the presence of mappings to a known vocabulary should allow it to derive some benefit).

  1. Concept identifiers should consist only of the letters a-z, A-Z, and numbers 0-9, i.e. no spaces, no exotic letters (for example umlauts), and no characters which would make a token inexpressible as part of a URI; since tokens are for use by computers only, this is not a big restriction, since the exotic letters can be used within the labels and documentation if appropriate.
  2. The concept identifiers should be kept in human-readable form, directly reflect the implied meaning, and not be semi-random identifiers only (for example, use spiralGalaxy, not t1234567); tokens should preferably be created via a direct conversion from the preferred label via removable/translation of non-token characters (see above) and sub-token separation via capitalisation of the first sub-token character (for example the label My favourite idea-label #42 is converted into MyFavouriteIdeaLabel42).
  3. Labels should be unchanged from the labels appearing in the source vocabulary. When developing a new vocabulary standard thesaurus practice indicates that english language labels for concrete concepts should be pluralised – thus "galaxies"@en, but "astronomy"@en; thesaurus practice in other european languages uses the singular for all cases.
  4. Each concept should have a definition (skos:definition) that constitutes a short description of the concept which could be adopted by an application using the vocabulary. Each concept should have additional documentation using SKOS Notes or Dublin Core terms as appropriate (see [std:skosref]). In practice, this requirement is rather difficult to satisfy, since pre-existing structured vocabularies, being convered to SKOS, frequently provide only labels, and not fuller descriptions or scope notes.
  5. The language localisation should be declared where appropriate, in preferred labels, alternate labels, definitions, and the like.
  6. Relationships (“broader”, “narrower”, “related”) between concepts should be present, but are not required; if used, they should be complete (thus all “broader” links have corresponding “narrower” links in the referenced entries and “related” entries link each other).
  7. “TopConcept” entries (see above) should be declared and normally consist of those concepts that do not have any “broader” relationships (i.e. not at a sub-ordinate position in the hierarchy).
  8. The SKOS standard describes some good practices for vocabulary maintenance, such as using <skos:changeNote> and the like, and these are elaborated in the (currently draft) note [kendall08]. Publishers should respect such good maintenance practices are are available.
  9. Publishers should publish “mappings” between their vocabularies and other commonly used vocabularies. These should be external to the defining vocabulary document so that the vocabulary can be used independently of the publisher's mappings. Open issue.

3.3. Good practices when serving vocabularies on the web [link here]

The W3C Interest Group Note Cool URIs for the Semantic Web [sauermann08] presents guidelines for the effective use of URIs when serving web documents and concepts on the Semantic Web. When providing vocabularies to the VO, we recommend that publishers conform to these guidelines in general. We make some further observations below.

The “Cool URIs” guidelines describe a number of desirable features of URIs in this context, namely simplicity, stability and manageability. Section 4.5 of the document describes these features as follows (quoted directly).

Simplicity
Short, mnemonic URIs will not break as easily when sent in emails and are in general easier to remember, e.g. when debugging your Semantic Web server.
Stability
Once you set up a URI to identify a certain resource, it should remain this way as long as possible. Think about the next ten years. Maybe twenty. Keep implementation-specific bits and pieces such as .php and .asp out of your URIs, you may want to change technologies later.
Manageability
Issue your URIs in a way that you can manage. One good practice is to include the current year in the URI path, so that you can change the URI-schema each year without breaking older URIs. Keeping all 303 URIs on a dedicated subdomain, e.g. http://id.example.com/alice, eases later migration of the URI-handling subsystem.

We endorse this advice in this Recommendation: VO vocabularies should use URIs which have these properties. The advice in the third point is a general point about maintaining the general URI namespace on a particular server, and is not about versioning vocabulary namespaces.

The “Cool URIs” document also describes two broad strategies for making these URIs available on the web, which they name 303 URIs and hash URIs (see the document, section 4, for descriptions). They note that the hash URI strategy “should be preferred for rather small and stable sets of resources that evolve together. The ideal case[s] are RDF Schema vocabularies and OWL ontologies, where the terms are often used together, and the number of terms is unlikely to grow out of control in the future.” Since this is the case for the (relatively small) SKOS vocabularies this Recommendation discusses, and since an application will generally want to use the complete vocabulary rather than only single concepts, we suggest that vocabularies conformant to this Recommendation should be distributed as hash URI ones.

Common to the two strategies above is the insistence that the vocabulary URIs are HTTP URIs which are retrievable on the web – they differ only in the practicalities of achieving this. The strategies also share the expectation that the vocabulary URIs are retrievable both as RDF (machine-readable) and as HTML (providing documentation for humans). We elevate this to a requirement of this Recommendation: vocabulary terms must be HTTP URIs which must be dereferenceable as both RDF and HTML using the mechanism appropriate to the URI naming strategy.

3.4. Example: serving the A&A vocabulary

While [sauermann08] discusses the design of the URIs naming concepts, it says little about the mechanics of making these available on the web. We refer vocabulary publishers to the recipe advice contained in [berrueta08], which we illustrate here in the case of the hash URI strategy.

The A&A vocabulary has the namespace http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf/AAkeys. In accordance with the above guidelines, this namespace URI is dereferenceable, and if you enter the URI into a web browser, you will end up at a page describing the vocabulary. The way this works can be illustrated by using curl to dereference the URI (URIs are cropped for legibility):

% curl --head http://[...]/rdf/AAkeys
HTTP/1.1 303 See Other
Date: Thu, 08 May 2008 14:07:12 GMT
Server: Apache
Location: http://[...]/rdf/vocabularies-2008-05-08/AAkeys/AAkeys.html
Connection: close
Content-Type: text/html; charset=iso-8859-1

The server has responded to the HTTP GET for the URI with a 303 response, and a Location header, pointing to the HTML representation of this thing.

If we instead request an RDF representation, by stating a desired MIME type in the HTTP Accept header, we get a slightly different response:

% curl --head -H accept:text/turtle http://[...]/rdf/AAkeys
HTTP/1.1 303 See Other
Date: Thu, 08 May 2008 14:11:28 GMT
Server: Apache
Location: http://[...]/rdf/vocabularies-2008-05-08/AAkeys/AAkeys.ttl
Connection: close
Content-Type: text/html; charset=iso-8859-1

This is also a 303 response, but the Location header this time points to an RDF file in Turtle syntax, which we can now retrieve normally.

% curl --include http://[...]/rdf/vocabularies-2008-05-08/AAkeys/AAkeys.ttl
HTTP/1.1 200 OK
Date: Thu, 08 May 2008 14:13:35 GMT
Server: Apache
Content-Type: text/turtle; charset=utf-8

@base <http://[...]/rdf/AAkeys> .
@prefix rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#> .
@prefix dc: <http://purl.org/dc/elements/1.1/> .
@prefix rdfs: <http://www.w3.org/2000/01/rdf-schema#> .
@prefix : <http://www.w3.org/2004/02/skos/core#> .

<>
    dc:created "2008-05-08" ;
    dc:title "Vocabulary for Astronomy & Astrophysics Journal keywords (Version wd-1.0)"@en ;
    a :ConceptScheme ;

# and so on...

Note that the base URI in the returned RDF still refers to the unversioned concept names.

This behaviour is controlled by (in this case) an Apache .htaccess file which looks like this:

AddType application/rdf+xml .rdf
# The MIME type for .n3 should be text/rdf+n3, not application/n3:
# see MIME notes at <http://www.w3.org/2000/10/swap/doc/changes.html>
#
# The MIME type for Turtle is text/turtle, though this has not
# completed its registration: see
# <http://www.w3.org/TeamSubmission/turtle/#sec-mediaReg>
AddType text/rdf+n3 .n3
AddType text/turtle .ttl
# For Charset types, see <http://www.iana.org/assignments/character-sets>
AddCharset UTF-8 .n3
AddCharset UTF-8 .ttl

RewriteEngine On
# This will match the directory where this file is located.
RewriteBase /users/norman/ivoa/vocabularies/rdf

RewriteCond %{HTTP_ACCEPT} application/rdf\+xml
RewriteRule ^(AAkeys|AOIM|UCD|IVOAT|IAUT93)$ vocabularies-2008-05-08/$1/$1.rdf [R=303]

RewriteCond %{HTTP_ACCEPT} text/rdf\+n3 [OR]
RewriteCond %{HTTP_ACCEPT} application/n3 [OR]
RewriteCond %{HTTP_ACCEPT} text/turtle
RewriteRule ^(AAkeys|AOIM|UCD|IVOAT|IAUT93)$ vocabularies-2008-05-08/$1/$1.ttl [R=303]

# No accept conditions: make the .html version the default
RewriteRule ^(AAkeys|AOIM|UCD|IVOAT|IAUT93)$ vocabularies-2008-05-08/$1/$1.html [R=303]

These various RewriteRule statements examine the content of the HTTP Accept header, and return 303-redirections to the appropriate actual resource.

Note that the namespace remains unversioned throughout the maintainance history of this vocabulary, even though the actual RDF files being returned might change as labels or relationships are adjusted. Previous versions of the vocabulary RDF will remain available, though they will no longer be served by dereferencing the namespace URL.

4. Example vocabularies (informative) [link here]

The intent of having the IVOA adopt SKOS as the preferred format for astronomical vocabularies is to encourage the creation and management of diverse vocabularies by competent astronomical groups, so that users of the VO and related resources can benefit directly and dynamically without the intervention of the IAU or IVOA. However, we felt it important to provide several examples of vocabularies in the SKOS format as part of the proposal, to illustrate their simplicity and power, and to provide an immediate vocabulary basis for VO applications.

The vocabularies described below are included, as SKOS files, in the distributed version of this standard. These vocabularies have stable URLs, and may be cited and used indefinitely. These vocabularies will not, however, be developed as part of the maintenance of this standard. Interested groups, within and outwith the IVOA, are encouraged to take these as a starting point and absorb them within existing processes.

The exceptions to this rule are the constellation vocabulary, provided here mainly for didactic purposes, and the proposed IVOA Thesaurus, which is being developed as a separate project and whose aim is to provide a corrected, more user-friendly, more complete, and updated version of the 1993 IAU thesaurus. Although work on the IVOA Thesaurus is on-going, the fact that it is largely based on the IAU thesaurus means that it is already a very useful resource, so a usable snapshot of this vocabulary will be published with the other examples.

We provide a set of SKOS files representing the vocabularies which have been developed, and mappings between them. These vocabularies have base URIs starting http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf, and can be downloaded at the URL

http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf/vocabularies-2008-05-08.tar.gz

4.1. A Constellation Name Vocabulary [link here]

This vocabulary is presented as a simple example of an astronomical vocabulary for a very particular purpose, such as handling constellation information like that commonly encountered in variable star research. For example, “SS Cygni” is a cataclysmic variable located in the constellation “Cygnus”. The name of the star uses the genitive form “Cygni”, but the alternate label “SS Cyg” uses the standard abbreviation “Cyg”. Given the constellation vocabulary, all of these forms are recorded together in a computer-manipulatable format. Various incorrect forms should probably be represented in SKOS hidden labels.

The <skos:ConceptScheme> contains a single <skos:TopConcept>, “constellation”

XML Syntax Turtle Syntax
<skos:Concept rdf:about="#constellation">
  <skos:inScheme rdf:resource=""/>
  <skos:prefLabel>
    constellation
  </skos:prefLabel>
  <skos:definition>
    IAU-sanctioned constellation names
  </skos:definition>
  <skos:narrower rdf:resource="#Andromeda"/>
  ...
  <skos:narrower rdf:resource="#Vulpecula"/>
</skos:Concept>

<#constellation> a :Concept;
  :inScheme <>;
  :prefLabel "constellation";
  :definition "IAU-sanctioned constellation names";
  :narrower <#Andromeda>;
  ...
  :narrower <#Vulpecula>.

and the entry for “Cygnus” is

<skos:Concept rdf:about="#Cygnus">
  <skos:inScheme rdf:resource=""/>
  <skos:prefLabel>Cygnus</skos:prefLabel>
  <skos:definition>Cygnus</skos:definition>
  <skos:altLabel>Cygni</skos:altLabel>
  <skos:altLabel>Cyg</skos:altLabel>
  <skos:broader rdf:resource="#constellation"/>
  <skos:scopeNote>
    Cygnus is nominative form; the alternative
    labels are the genitive and short forms
  </skos:scopeNote>
</skos:Concept>

<#Cygnus> a :Concept;
  :inScheme <>;
  :prefLabel "Cygnus";
  :definition "Cygnus";
  :altLabel "Cygni";
  :altLabel "Cyg";
  :broader <#constellation>;
  :scopeNote """Cygnus is nominative form;
    the alternative labels are the genitive and
    short forms""" .

Note that SKOS alone does not permit the distinct differentiation of genitive forms and abbreviations, but the use of alternate labels is more than adequate enough for processing by VO applications where the difference between “SS Cygni”, “SS Cyg”, and the incorrect form “SS Cygnus” is probably irrelevant.

4.2. The Astronomy & Astrophysics Keyword List [link here]

Namespace: http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf/AAkeys.

This vocabulary is a set of keywords maintained jointly by the publishers of the journals Astronomy and Astrophysics (A&A), Monthly Notices of the Royal Astronomical Society (MNRAS) and the Astrophysical Journal (ApJ). As noted in the introduction, an analysis of these keywords [preitemartinez07] indicates that the different journals are slightly inconsistent with each other; we have rather arbitrarily used the list from the A&A web site. The intended usage of the vocabulary is to tag articles with descriptive keywords to aid searching for articles on a particular topic.

The keywords are organised into categories which have been modelled as hierarchical relationships. Additionally, some of the keywords are grouped into collections which has been mirrored in the SKOS version. The vocabulary contains no definitions or related links as these are not provided in the original keyword list, and only a handful of alternative labels and scope notes that are present in the original keyword list.

4.3. The AOIM Taxonomy [link here]

Namespace: http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf/AOIM.

This vocabulary is published by the IVOA to allow images to be tagged with keywords that are relevant for the public. It consists of a set of keywords organised into an enumerated hierarchical structure. Each term consists of a taxonomic number and a label. There are no definitions, scope notes, or cross references.

When converting the AOIM into SKOS, it was decided to model the taxonomic number as an alternative label. Since there are duplication of terms, the token for a term consists of the full hierarchical location of the term. Thus, it is possible to distinguish between

Planet -> Feature -> Surface -> Canyon

and

Planet -> Satellite -> Feature -> Surface -> Canyon

which have the tokens PlanetFeatureSurfaceCanyon and PlanetSatelliteFeatureSurfaceCanyon respectively.

4.4. The UCD1+ Vocabulary [link here]

Namespace: http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf/UCD.

The UCD standard is an officially sanctioned and managed vocabulary of the IVOA. The normative document is a simple text file containing entries consisting of tokens (for example em.IR), a short description, and usage information (“syntax codes” which permit UCD tokens to be concatenated). The form of the tokens implies a natural hierarchy: em.IR.8-15um is obviously a narrower term than em.IR, which in turn is narrower than em.

Given the structure of the UCD1+ vocabulary, the natural translation to SKOS consists of preferred labels equal to the original tokens (the UCD1 words include dashes and periods), vocabulary tokens created using guidelines in 3.2. Good practices of vocabulary design (for example, "emIR815Um" for em.IR.8-15um), direct use of the definitions, and the syntax codes placed in usage documentation: <skos:scopeNote>UCD syntax code: P</skos:scopeNote>

Note that the SKOS document containing the UCD1+ vocabulary does NOT consistute the official version: the normative document is still the text list. However, on the long term, the IVOA may decide to make the SKOS version normative, since the SKOS version contains all of the information contained in the original text document but has the advantage of being in a standard format easily read and used by any application on the semantic web whilst still being usable in the current ways.

4.5. The 1993 IAU Thesaurus [link here]

Namespace: http://www.astro.gla.ac.uk/users/norman/ivoa/vocabularies/rdf/IAUT93.

The IAU Thesaurus consists of concepts with mostly capitalised labels and a rich set of thesaurus relationships (“BT” for "broader term", “NT” for “narrower term”, and “RT” for “related term”). The thesaurus also contains “U” (for “use”) and “UF” (“use for”) relationships. In a SKOS model of a vocabulary these are captured as alternative labels. A separate document contains translations of the vocabulary terms in five languages: English, French, German, Italian, and Spanish. Enumerable concepts are plural (for example “SPIRAL GALAXIES”) and non-enumerable concepts are singular (for example “STABILITY”). Finally, there are some usage hints like “combine with other”, which have been modelled as scope notes.

In converting the IAU Thesaurus to SKOS, we have been as faithful as possible to the original format of the thesaurus. Thus, preferred labels have been kept in their uppercase format.

The IAU Thesaurus has been unmaintained since its initial production in 1993; it is therefore significantly out of date in places. This vocabulary is published for the sake of completeness, and to make the link between the evolving vocabulary work and any uses of the 1993 vocabulary which come to light. We do not expect to make any future maintenance changes to this vocabulary, and would expect the IVOAT vocabulary, based on this one, to be used instead (see 4.6. Towards an IVOA Thesaurus).

4.6. Towards an IVOA Thesaurus [link here]

While it is true that the adoption of SKOS will make it easy to publish and access different astronomical vocabularies, the fact is that there is no vocabulary which makes it easy to jump-start the use of vocabularies in generic astrophysical VO applications: each of the previously developed vocabularies has their own limits and biases. For example, the IAU Thesaurus provides a large number of entries, copious relationships, and translations to four other languages, but there are no definitions, many concepts are now only useful for historical purposes (for example many photographic or historical instrument entries), some of the relationships are false or outdated, and many important or newer concepts and their common abbreviations are missing.

Despite its faults, the IAU Thesaurus constitutes a very extensive vocabulary which could easily serve as the basis vocabulary once we have removed its most egregious faults and extended it to cover the most obvious semantic holes. To this end, a heavily revised IAU thesaurus is in preparation for use within the IVOA and other astronomical contexts. The goal is to provide a general vocabulary foundation to which other, more specialised, vocabularies can be added as needed, and to provide a good “lingua franca” for the creation of vocabulary mappings.

5. Mapping vocabularies (informative) [link here]

Part of the motivation for formalising vocabularies within the VO is to support mapping between vocabularies, so that an application which understands, or can natively process, one vocabulary, can use a mapping to provide at least partial support for data described using another vocabulary. The SKOS document ...is still a draft, and may or may not end up with support for mappings in the core document rather than in a companion document.

Example mappings to appear: see issue [mappings-6].

Appendices

References [link here]

[berrueta08] Diego Berrueta and Jon Phipps.
Best practice recipes for publishing RDF vocabularies. W3C Working Draft, January 2008. [Online].
[dc:namespaces] Andy Powell and Harry Wagner, editors.
Namespace policy for the dublin core metadata initiative (DCMI). DCMI Recommendation, July 2007. [Online].
[hessman05] F V Hessman.
VOConcepts — a proposed UCD for astronomical objects, events, and processes. [Online, cited January 2008].
[kendall08] Elisa Kendall and Vit Novacek, editors.
Principles of good practice for managing RDF vocabularies and OWL ontologies. W3C Editor's Draft, March 2008. [Online].
[lortet94] M-C Lortet, S Borde, and F Ochsenbein.
Second reference dictionary of the nomenclature of celestial objects. Astron. Ap. Supp, 107 pp. 193-218, 1994. [Online].
[lortet94a] M-C Lortet, S Borde, and F Ochsenbein.
The second reference dictionary of the nomenclature of celestial objects (solar system excluded). volumes i, ii.. Technical Report 24, Centre de Données astronomique de Strasbourg, 1994. [Online].
[notation3] Tim Berners-Lee.
Notation 3. Web page, 2006. [Online, cited 2008 May 8].
[preitemartinez07] Andrea Preite Martinez and Soizick Lesteven.
Astronomical keywords in the era of the virtual observatory. IVOA Note, IVOA, 2007. [Online].
[sauermann08] Leo Sauermann and Richard Cyganiak.
Cool URIs for the semantic web. W3C Interest Group Note, March 2008. [Online].
[shobbrook92] R.M. Shobbrook and R.R. Shobbrook.
The IAU thesaurus for improved on-line access to information. Proc. Astron. Soc. of Australia, 10 pp. 134, 1992. [Online].
[std:aoim] Robert Hurt, Lars Lindberg Christensen, and Adrienne Gauthier.
Astronomical outreach imagery metadata tags for the virtual observatory. [Online, cited January 2008].
[std:bs8723-1] Structured vocabularies for information retrieval — guide — definitions, symbols and abbreviations (BS 8723-1:2005).
British Standard, 2005.
[std:dublincore] DCMI Usage Board.
DCMI metadata terms. DCMI Recommendation, 2006. [Online].
[std:iso5964] Documentation — guidelines for the establishment and development of multilingual thesauri (ISO 5964:1985=BS 6723:1985).
International Standard, 1985.
[std:ivoa-astro-onto] Sébastien Derriere, Andrea Preite Martinez, and Alexandre Richard, editors.
Ontology of astronomical object types. IVOA Working Draft, 2007. [Online].
[std:owl] World Wide Web Consortium.
The web ontology language. [Online].
[std:rdfxml] Dave Beckett.
RDF/XML syntax specification (revised). W3C Recommendation, February 2004. [Online].
[std:rfc2119] S Bradner.
Key words for use in RFCs to indicate requirement levels. RFC 2119, March 1997. [Online].
[std:skosref] Alistair Miles and Sean Bechhofer, editors.
SKOS reference. W3C Working Draft, January 2008. [Online].
[std:turtle] Dave Beckett.
Turtle - terse RDF triple language. W3C Team Submission, January 2008. [Online].
[std:ucd] Sébastien Derriere, Andrea Preite Martinez, and Roy Williams, editors.
UCD (Unified Content Descriptor) — moving to UCD1+. IVOA Recommendation, 2004. [Online, cited February 2008].
[std:voevent] Rob Seaman and Roy Williams, editors.
Sky event reporting metadata (VOEvent). IVOA Recommendation, 2006. [Online].
[std:z39.19] Guidelines for the construction, format and management of monolingual thesauri (ANSI/NISO Z39.19-2005).
American National Standard, 2005. Closely corresponds to BS 8723:2005, parts 1 and 2, which replaces BS 5723; and to forthcoming ISO 25964, which replaces ISO 2788. [Online].

Revision: 420 Date: 2008-05-08 16:25:14 +0100 (Thu, 08 May 2008)