Attendees
- Bruce Herbert, TAMU
- Daniel Fuentes, University of Grenada, Spain
- Graham Triggs, Duraspace
- John Mark Ockerbloom, UPenn
- Manuel Hidalgo, ICAA, Costa Rica
- Noella Natalino, Plum Analytics
- Ralph O'Flinn, University of Alabama, Burmingham
- Sukie, Dartmouth College
- Eric Meeks, UCSF
- Mike Conlon, Duraspace
- Jim Blake, Cornell
- Tina Delshaw, Stanford
- Muhammad Javed, Cornell
Announcements
Conference next week http://vivoconference.org
- VIVO 1.10 Beta 2 before conference. Release candidate and release after conference.
Notes
- Mike showed the TPF server provided in VIVO 1.10. It is currently available in OpenVIVO at http://openvivo.org/tpf/core Very easy to write code to fetch triples. The responses are very fast due to the restriction to pattern matching triples. Might be a basis for a federated search of sites, the number one request of VIVO sites for many years. Mike also showed preliminary Javascript code that can be used to query a TPF server. Very simple. No SPARQL necessary. Open API. Very fast. Will still face speed limitations for some applications. Smart design necessary as always.
- Discussion of the global architecture for federated search. VIVO sites can participate directly. Data sites such as ResearchGraph (Neo4J based) can participate by adding a TPF server to a triple store – mapping the ResearchGraph structures to VIVO has been completed. Similarly for ORCiD, WikiCite and other data sources. Still others can participate through an aggregator such as Dave Eichmann's CTSAsearch – a TPF server there would provide access to about 100 sites worth of data.
- Federated search may well require improved use of globally identifiers to reconcile identity. OpenVIVO provides a presumptive view of this approach. Everything in OpenVIVO is identified using a global identifier: Organizations by GridId, People by ORCiD, works by DOI and PMID, Journals by ISSN.
- Mike will share the code after some minor improvements to support the general case of VIVO retrieval, including arbitrary length "paths" to data – the current implementation is limited to path length two: entity predicate entity literal. Some VIVO data requires paths of the form entity predicate entity predicate entity literal. Will also bundle up the retrievals into simply named functions to hide ontological detail.
