Semantic Web Best Practices and Deployment

RDF/Topic Maps Task Force

Description of Work

Last updated: $Date: 2004/12/02 14:31:02 $


RDF/Topic Maps Task Force




Steve Pepper, Ontopia


  • Lars Marius Garshol, Ontopia
  • Nicola Gessa, University of Bologna
  • Marco Pirruccio, University of Bologna
  • Valentina Presutti, University of Bologna
  • Bernard Vatant, Mondeca
  • Fabio Vitali, University of Bologna


The long-term objective of this task force is to promote interoperability between the W3C's RDF/OWL family of specifications and the ISO's family of Topic Maps standards.

The mapping between the two architectures:

The immediate short-term objective is to create a framework for defining mappings between the Topic Maps Data Model and the core RDF model such that applications can "view" Topic Maps as though it were RDF data and vice versa.

More specific longer-term objectives are likely to be:

  • Defining mappings between RDFS and Topic Maps
  • Defining mappings between OWL and Topic Maps Constraint Language (TMCL)


  1. Create an overview of already-published proposals for RDF/TM mappings Ontopia: A single sentence (5-6 lines) detailing what we mean by vocabulary-based conversion and schema-based conversion
  2. Requirements: A list of 3- terms well-explained to identify the overall constraints of the translation mechanism we are going to propose
      • Completeness: Translation should cover all semantic aspects of RDF and TM models (as defined in their respective specifications)
      • Readability: The converted document should be as close as possible to what a human expert would have created if he/she wrote the specification in the target language
      • Reversability: To convert a document the target languages creates a document that is as close as possible to the original (the same graph in RDF, the same abstract document in TM)

    defining a conversion mechanism between the RDF and Topic Maps paradigms to enhance their interoperability, three basic requirements have to be satisfied by a really useful translation framework:

      • Completeness: RDF and Topic Maps adopt different semantic models to express meta-data, so a conversion schema should maintain the whole semantic content of the original assertions, regardless of the syntax or the XML serialisation used
      • Readability: converting a document from one paradigm to the other should lead to a result that has to be as much as possible similar to a specification written from scratch by an human expert in the target language
      • Reversibility: the conversion procedure has to be completely reversible: a round-trip translation process should lead to the original document, both starting from an RDF or Topic Maps definition.

  3. Choose one or more of these as a starting point for defining a complete methodology Ontopia: A single sentence justifying why the Ontopia-UniBo method has better chances to reach a good translation than the others
      • First proposal: Ontopia - UniBo?
      • Why not the other ones?
        • Ogievetsky: add basic problems
        • Lacher, Decker: add basic problems
        • Moore: add basic problems
        • Prud'hommeaux: add basic problems
        • van der Vlist: add basic problems

  4. Main challenges in the conversion Ontopia: A single sentence for each of these items
    • Identifiers
    • variant names
    • Scope
    • N-ary associations
    • Reification
    • Roles

  5. Overview of the translation model Unibo: A few sentences (10-15 lines) detailing the following list
    1. Define a standard fallback mechanism for translation that relaxes the "readability" constraints but maintains a guaranteed 100% determinism. -- This might be unnecessary if level 1 is found to be deterministic
    2. Define a standard (quite readable) translation model based on deductions on the usage of terms, e.g.:
      • an RDF property is an occurrence if the term is a literal
      • an RDF property is an association if the term is a URL.
    3. Allow for schema languages (RDFS, TMCL, OWL) to suggest hints for natural translation, e.g.:
      • Making use of rdfs:range and rdfs:domain
      • Making use of OWL properties such as owl:inverseOf (with drawing)
    4. Allow for humans to suggest hints for natural translation (in an XML dialect), e.g.:
      • "author" is an association
      • "birth date" is an occurrence

    The proposed translation model can be laid out in different levels. Each level satisfies the overall contraints of the translation mechanism as mentioned before.

    The first level defines a standard fallback mechanism for translation in absence of any information about the domain of the metadata. In this case it might be necessary to relax the readability constraints. At this level we do not care about the meaning of the terms to translate: we might end outputting a document quite different from a document written in the target language, since we produce many abstract resources (RDF resources and topics) to mantain all the data and metadata of the original document.

    This mechanism is meant to translate each term in the most general way. For example:

    • a binary Topic Maps association is translated as an n-ary association;
    • a baseName element produces all the abstract resources necessary to maintain the information of its scope and variant children.

    The second level defines a standard translation model based on deductions on the usage of terms. For example:

    • an RDF property is an occurrence if the object is a literal;
    • an RDF property in an association if the object is an addressable resource;
    • a Topic Maps baseName with only a baseNameString is a rdfs:label.

    The output document has quite a readable syntax but its content is defined only by the usage of the terms of the source document.

    The third level, whenever possible, leverages the RDFS and TMCL schemas or any other higher level language such as OWL in order to obtain hints for the translation. For example:

    • if the RDFS schema uses rdfs:range and rdfs:domain to define the predicates, it is possible to know if they have to be translated as associations or occurrences;
    • Topic types, association types, occurrence types, superclass-subclass associations, etc. are expressed with TMCL and PSIs. In this way it is possible to know, for example, which occurrence types or association types will be translated as rdfs:Predicates.

    A fourth level might become necessary if no schema is available or if a more detailed (or anyway different) translation is necessary than what is produced by the previous levels. In this case it is possible to suggest hints for a natural translation. This level lets domain experts express these hints in an metadata resource that specifies how to translate the terms. For example it is possible to express that:

    • the associations with association-type "author-of" will be translated as an RDF property among resources;
    • the occurrence with occurrence-role-type "birthdate" will be translated as an RDF property whose rdfs:range is a literal;
    • the topics with association-role-type "author" and "website" will be translated as the domain and range of the RDF property "author-of".

  6. Identify lacunae and weaknesses in the chosen approach Ontopia: A one line CYA sentence detailing that the proposed solution NEED to be detailed and discussed
  7. Write and publish a Best Practices recommendation describing the methodology and vocabulary Whoever: One sentence (2 lines) summarizing what is the planned output of the TF (i.e., the content of the deliverables section)


The Task Force is chartered in the first instance to cover the interpretation of basic RDF data as Topic Maps and basic Topic Maps as RDF. Every aspect of the model described in the core RDF specifications is in scope, as is every aspect of the model described in the core Topic Maps standards.

Translation of anything beyond and above the level of data (as expressed in RDF and TM) is out of scope. In particular, translation of concepts relating to OWL, TMCL, TMQL are out of scope in the first instance. However it is envisaged that the Task Force may be re-chartered after finalizing its first set of deliverables in order to tackle some of these more advanced issues.


Unibo: two lines for each of the documents in the list specifying what will be inside, the relationship with the other documents and the overall importance within the activities of the TF

Description of expected results:

  1. Initial note summarizing the current state of the art
  2. Guidelines for basic RDF/TM interoperability
  3. Vocabulary for mapping RDF properties to Topic Maps (could be an appendix to 2.)
  4. Vocabulary for mapping Topic Maps properties to RDF (could be an appendix to 2. - could be the same vocabulary as 3.)
The output of this TF is expected to provide the fundamental guidelines for the correct and fruitful interoperability of RDF and Topic Maps. This will take the form of two main documents and two additional resources (which could even be considered as appendices to one of the main documents):

  1. Survey: an overview of the current efforts in translation and conversion of metadata sets between the two languages, both from academia and industry. Delivery date: 4 months after start.
  2. Conversion Guidelines: The main document for ensuring interoperability between RDF and Topic Maps, providing background and justifications for the four levels of conversion previously listed, as well as details on the conversion model suggested. Dellivery date: 8 months after start.
  3. RDF-TM Vocabulary: A list of rules for mapping RDF properties into Topic Maps, as needed by levels 3 and 4 of the previously illustrated model. This document could be in fact an appendix of document 2.
  4. TM-RDF Vocabulary: A list of rules for mapping Topic maps into RDF properties, as needed by levels 3 and 4 of the previously illustrated model. This document could be in fact an appendix of document 2, and hopefully could even coincide document 3.


Ontopia: Expand the following sentence

Information owners and application developers who wish to integrate RDF and Topic Maps data.

  • Use cases Both: Expand in 3 lines each of these use cases stressing the issues that need to be solved in terms of relationships between RDF and TM
    • A government portal dealing with metadata coming from different sources in both languages and needing to convert them into one languages to infer new knowledge, and thus needing to reconciliate metadata schema coming from different sources.
    • A TM application needing to use RDF-based web services to perform tasks.
    • A TM application needing to export its data in RSS format
    • A semantic search engine crawling the Web and the Semantic Web should be able to exploit all the significant meta-information attached to the resources it indexes. The ranking process of such a semantic search engine would expect as its input a (uniform) RDF-based set of metadata together with the associated content. A resource which is part of a TM-based information source should be indexed just as it is for the RDF-based ones. A TM-to-RDF mapping manager would support this task.
    • An organization that needs to create its own RDF-ontology, which finds (or needs to share) an existing TM-based information source that supports a wide part of its needs. Such organization may use the RDF mapping of the existing TM-based information source as a starting set of concepts and relations, in order to define the ontology.

    DEPENDENCIES (if any)

    • New version of TM standard
    • New version of RDF standard
    • (in particular): Defining N-ary Relations on the Semantic Web: Use With Individuals, W3C Working Draft 21 July 2004,
    • Other TFs within SWBPD and other W3C working groups about Semantic Web. E.g.: vocabularies?
    • (Lesser): Public drafts or errata of TMQL, TMCL, RDFS, OWL coming out in the next months

      N.B.: We expect that none of these drafts have an actual impact on the translation model. On the other hand, we will definitely consider the issue



    Other bibliography worth mentioning but unrelated

    Integrating Topic Maps and OWL

    Definitely not yet in scope

    Regarding Identifiers in Topic Maps and RDF

    They are not proposals for whole mapping but solving special cases. Thus partially in scope but worth being reviewd at a later date.

    -- FabioVitali - 19 Oct 2004

    -- FabioVitali - 28 Nov 2004

    Revision: r1.3 - 02 Dec 2004 - 15:31 - StevePepper
    RDFTM > SurveyOfExistingProposals > SomeInitialDocuments
    Copyright © 1999-2017 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
    Ideas, requests, problems regarding Fabio's Wiki? Send feedback