Applied Linked Data Call 2015-04-15

Introductions / Attendees:

Steven Anderson -- Interested in moving to LDP & Linked Data. 
Trey Terrell -- Oregon Digital. Hydra runs all linked data 
Corey Harper-- Interested, particularly SOLR &  
Arwen Hutt -- Metadata Focus. Everyone encouraged her to stay on 
Aaron Coburn -- DOing stuff with linked data, need to know how to do it 
Andrew Woods -- LDP for Fedora. Also interested in PCDM 
Rob Anderson -- Linked Data & Standards. Member of W3C LDP Group  
 
 Is this about LDP or linked data in general? 
 
 What do we mean by LDP? The spec?    
  • Is this about using Linked Data in Hydra?    
  • Other names? Applied Linked Data? Linked Data Apps
  • People liked Applied LD. Need to run by Karen E
Linked Data Fragments Report-out
  • Triple Pattern Fragments as basic subset
  • Matching Triples for S-P-O
  • Keyword search
  • LDF Client, can run query against interface & do sparql filters client side
  • Explicit ranking of results to a matched keyword to allow for aggregation of linked data results
  • Next steps of building a Ruby interface for this (potentially in projecthydra-labs)
  • LDP Paging in header spec vs. LDF having a "next page" in the data
  • Rob's point that this break's the distinction between the data & metadata
  • From practical point of view, we'd need to determine which paging spec to adopt
  • Because we're already using LDP, it makes sense to do paging in the LDP way.
  • Though arguably, LDP & LDF serve different potential purposes. Restful Repo vs. query UI  [RS: +1 to this distinction if the two can be kept separate]
Solr
  • Getting not only the primary label for faceting but also having the alt label available for searching and such,
  • Idea of linked data as a discovery tool as well as display
  • Talk of using a messanging system when once your label is updated in your Linked Data interface, it runs a background job to update all of the objects in the system.
  • No atomic updates in SOLR, so you need to reindex the whole object.  That's the bottleneck.
Questions of pesistence. 
  • Storing a local mirror; local proxies with local URIs owl:sameAs'd to the source
  • But you don't necessarily want to "own" triples that you dont' really have ownership for
  • This is potentially a preservation question
  • Cache triple with subject id.loc and cache label from id.loc autocomplete api. Keep cached predicate for fall-back
  • Out of scope
How to do this solr stuff:
  • Faceting & cross references
  • LDF that caches?
  • Builds on Oregon Digital or Linked Vocabs?
  • Sidecar triplestore; Oregons use MongoDB, hoping to move to Marmota
  • Aaron recommends Stanbol
  • Trey asks how it works:
  • Takes rdf
  • indexes internally in Jena
  • Those triples are indexed in "SOLR Yards"
  • Get back JSON-LD