[seqfan] Linked Open Data (LOD) and the OEIS

Marc LeBrun mlb at well.com
Thu Mar 1 19:08:02 CET 2012

I was reminded again recently that the OEIS would be an ideal database to
contribute to the world's knowledge in the form of "Linked Open Data" (LOD).

You can easily find more about LOD by searching the web; an introduction is
here: http://en.wikipedia.org/wiki/Linked_data

Briefly, LOD is a foundation for the "semantic web", and is actively being
advocated by Tim Berners-Lee, the W3C and others.

Wikipedia notes the LOD available on the web grew from 2 billion triples in
2009 to 31 billion in 2011.  There are many domains represented, including
medicine, law, media, science and others.  The OEIS would be a natural fit.

There are powerful technologies available for using LOD and more will be
coming.  These would complement the OEIS "native" mathematical strengths
(such as lookup and superseeker) to help address all the OTHER valuable
"metadata" that is included in each entry.

Some amazing applications have already been built federating data from
different domains.   For example crossing pharmaceutical info and medical
research.  The OEIS would naturally federate with things like publication
citation databases, and so on.

LOD is a hot topic these days, and I believe the OEISF could perhaps even
get some support to help it join the semantic web community.

Of course many of us are very busy, but I think this is an exciting idea, so
if you are personally interested in pursuing LOD for the OEIS further please
let me know (off list) and maybe we can figure out a good way to collaborate
on it somehow...



To make this more concrete and relevant here's a brief description of what
LOD would be like from the perspective of the OEIS:

The basis of LOD is to reference everything with HTTP URIs.  Of course all
the OEIS entries are already identified in this way.  For example A007317,
"Binomial transform of Catalan numbers", has the URL
http://oeis.org/A007317 and so forth.

Building on that URI foundation, all other knowledge is represented as
triples (implemented in RDF) that describe relationships between entities.

For example some (informal) triples related to A007317 would include:

  A007317  --hasName-->  "Binomial transform of Catalan numbers"

  A007317  --binomialTransform--> A000108

  A000108  --isCalled-->  "Catalan numbers"

  A000108  --isCalled-->  "Segner numbers"

Similarly other entities can be ID'd, say http://oeis.org/user/Marc+LeBrun
and used (eg in "hasAuthor" triples), related to eachother and so on.

The triple "schema" is itself also represented as triples, and thus is used
to provide an ontology of the domain:

  binomialTransform  --hasInverse-->  inverseBinomialTransform

This allows automatic reasoners to make useful deductions like

  A000108  --inverseBinomialTransform-->  A007317

and many more complicated inferences, removing the need to explicitly
represent them in the database itself.

By use of namespaces any number of ontologies can be used together, which
enables sharing existing standard ontologies.  Only truly OEIS-oriented
relationships would need to be modeled.

This work seems like it could qualify as fundable research that would
produce a valuable public resource.

More information about the SeqFan mailing list