Semantic interoperability in a federated Crystallography information service Traugott Koch, UKOLN Presentation at "Digital repositories supporting eResearch: exploring the eCrystals Federation Model. EBank/R4L/SPECTRa Joint Consultation Workshop, London, 2006-10-20

1 Three areas of semantic interoperability

• Need to enhance interoperability between data, images, text and metadata (both open and proprietary, free and fee-based) in:
• Data repositories (institutional, disciplinary, national, international)
• Publication repositories
• Data and publications on the Internet
• Aggregators
• Databases
• Services
• New requirement and user studies needed with transition to federated service, to define the scope, the levels of the interoperability, the functionality and user interface solutions
• Semantic vs. syntactic interoperability:
Syntactic interoperability is about applying common formats and protocols for data transfer and merging (e.g. CIF, XML. OAI, Z39.50, SRW).
Semantic interoperability is about shared meaning of the content.
• Approaches to enhance semantic interoperability:
• Agreed common standards at all sites
• Conversion/normalization
• Mappings between site-specific solutions
• Metadata enhancement (vocabularies, schemes, mapping, names)
When and where this is done depends on the selected architecture
• For full semantic interoperability, all three areas need to be addressed:
• A Data structures (metadata profiles)
• B Categorial data (topics, classification)
• C Factual data (names, formulae, other named entities)

2 Data structures: metadata engineering

• Data model. eBank L
• Metadata Application Profile. eBank. L
• Namespaces. eBank terms. L
• XML Schema specification for eBank terms
• Formats for the OAI harvesting protocol. eBank METS. oai_dc.
• Data dictionaries, as in CIF/the Crystallographic Information Framework

Potential actions:

• Agree on metadata solution incl. value encoding for a future service
• Develop common Application Profile for Crystallography data
• Define different degrees of interoperability/adherence to common model, incl. a minimum level
• Take steps towards harmonization with the profiles of related publication servers

3 Categorial data: subject access

• Used in eBank: keywords adapted from IUCr World Directory of Crystallographers
• Used in eBank: "Compound class", only useful to filter searches
• Keywords used in Acta Cryst. A, B and in J. Appl. Cryst. (all IUCr)
• Keywords and classification used in the 25 abstract & indexing databases which cover Crystallography Jourals Online
• Other sources of topical crystallography vocabulary?

Potential actions:

• Clarify the data model question: what do the keywords characterize: elements of the data or usage/problem context?
• Develop and maintain common controlled keyword system and classification for Crystallography, with coverage of data-related topics
• Develop and maintain mapping to related discipline and generic classifications

4 Factual data: named entity standards/authorities

• Author and institutional names
• Uncontrolled and unvalidated
• World Directory of Crystallographers and the IUCr ID
• Library and university related name authority projects; national projects
• "Names" of the objects of study and their components: crystal structures, chemical compounds
• IUPAC Chemical Names (Colour books)
• InChI (International Chemical Identifier)
• Chemical Formula (based on CIF)
• Names used by others in the future federation?

Potential actions:

• Cooperate in building and improvement of name authority databases
• Contribute to further standardisation (InChI, Gold Book etc.)
• Use name authorities for verification and metadata enhancement
• Build authorities into metadata creation tools (Repository submission toolset. DSpace, eprint), e.g. via web services
• Experiment with mash-ups of crystallography data. What would be candidate reference schemes? Consequences for repositories?

5 Common standardisation issues

• To what a degree can existing conventions be seen as standard?
• Does the convention/standard lead consistently to the same "name"?
• Use of proprietary systems such as CAS numbers?
• General problems with common and standardised solutions: standardisation processes; adoption; validation; maintenance.

6 Application issues

• Metadata creation and subject assignment
• Same or similar tools as the ones supporting discovery (e.g. semi-automated indexing)
• "Cataloguing" rules in accordance with the data model and the Application Profile. Harmonization
• Explore benefits of text and data markup (e.g. CML) in combination with keyword indexing and fulltext-searching
• Experiments with participatory indexing (social tagging, folksonomies) carried out by research groups in specified research areas
• Investigate the need for formal ontologies and logical reasoning over the data and the literature in Crystallography
• Discovery: searching, browsing, linking
• Searching in components and substructures, using strings; graphical search support (JChempaint, Marvin applet) etc.


• Searching of (and filtering with) key characteristics of crystal structures (cf. Reciprocal Net, Crystallography Open Database)


• Searching inside the data files and the CIF format


• Searching in information services with much broader topical coverage (university-wide repositories, national data archives, OAIster, Google)


• Investigate the potential benefits of text and data mining for indexing and searching support


• Steps towards knowledge extraction, hypothesis creation and larger-scale computational processing