sTeX support for flexible metadata (MOLE)

[kohlhase]

We have published the idea that we can use OMDoc (and thus sTeX) for ontologies. In the Flexiform paper and the FormalSafe? case study we describe sTeX as a vocabulary specification language writing OMDoc1.3 flexible metadata annotations conveniently. Right now, we can only do this by writing custom *.sty and *.ltxml files for external packages.

It should be possible to internalize this into sTeX, just as \symdef internalizes the creation of semantic symbol macros into sTeX and \termdef for semantic concept macros.

This would allow us to bring sTeX light to a whole new level.

[kohlhase]

I guess the most important thing here is to decide on the interface we want to support here. \symdef introduces a macro for mathematical symbols that is valid only in Math mode. \termdef introduces a macro for \termref-cross-referenced technical terms. The natural thing for metadata relations are keys in the KeyVal? arguments to the sTeX macros and environments.

So I guess, we want to have a \keydef macro that works like \symdef, but introduces a new key to the metadata groups. We can build this using the omd infrastructure.

@Christoph, @Andrea are all metadata relations we are interested in expressible as single keys?

Otherwise we need a more elaborate interface.

Also we need a good (very simple) example. I guess the DC metadata should work for now.

[clange]

Replying to kohlhase:

So I guess, we want to have a \keydef macro that works like \symdef, but introduces a new key to the metadata groups. We can build this using the omd infrastructure.

What is omd anyway? I thought that is the start for “sTeX's support of flexible metadata”. The only thing that's missing from there is the mapping of keys to URIs for the output.

@Christoph, @Andrea are all metadata relations we are interested in expressible as single keys?

In principle yes. The problem is with the values:

• we need to allow URIs as values, and they must be distinguished from strings, e.g. DCsource=\uri{http://de.wikipedia.org/...}
• support for datatypes and language would also be nice. The datatype of a metadata property can be declared with the keydef, whereas the language can only be given for any concrete metadata field, maybe as an optional prefix, e.g. description=[de]irgendwas,...
• for the maximum of what we want to express (but maybe not using this interface), we need values that themselves are resources and can have their own metadata.

Also we need a good (very simple) example. I guess the DC metadata should work for now.

Maybe take part of the example I worked out for OMDoc. (The revision log using DCMI Terms) On the other hand that might be too complex (see above).

[clange]

Replying to clange:

Replying to kohlhase:

So I guess, we want to have a \keydef macro that works like \symdef, but introduces a new key to the metadata groups. We can build this using the omd infrastructure.

The only thing that's missing from there is the mapping of keys to URIs for the output.

Maybe actually not (only) URIs. We could refer to vocabulary terms by URI, e.g. mapping the knows key to http://xmlns.com/foaf/0.1/knows, or we could refer to them by symdef ID, if we think of metadata vocabularies/ontologies written in sTeX. For the former case of reusing existing metadata vocabularies (e.g. FOAF), we have two choices: either referencing them by URI, or writing sTeX module stubs (like those empty OMDoc theories that only have the semWebBase metadatum pointing to the URI of the “semantic web ontology namespace”)

[clange]

I just got the idea that we might not only want to create OMDoc+RDFa with sTeX as an input language, but maybe also XHTML+RDFa or any other XML+RDFa. Therefore it would make sense to base the implementation of this feature on OMDoc as little as possible. Which means:

• don't necessarily output OMDoc's <metadata> wrapper element (which we'll make optional in 1.3 anyway)
• maybe make the elements that carry RDFa attributes customizable/parametrizable, i.e. that when having an OMDoc output, one can say "when any RDFa triples are given for this element, create link and meta children and add them there", but that for XHTML output one can say "use div or span"

Still, for XHTML output we can do without that feature by creating slim OMDoc without any mathematical structure, just metadata, and postprocessing the OMDoc to XHTML with JOMDoc. But on the other hand we are good at producing XHTML from LaTeX (→ arXMLiv), so why not produce XHTML+RDFa from LaTeX with sTeX 1.3 metadata?