Date: Fri, 29 Mar 2024 07:54:08 -0400 (EDT)
Message-ID: <463868803.30334.1711713248054@lyrasis1-roc-mp1>
Subject: Exported From Confluence
MIME-Version: 1.0
Content-Type: multipart/related;
boundary="----=_Part_30333_967017012.1711713248053"
------=_Part_30333_967017012.1711713248053
Content-Type: text/html; charset=UTF-8
Content-Transfer-Encoding: quoted-printable
Content-Location: file:///C:/exported.html
2015-02-24 breakout: Services on linked data
2015-02-24 breakout: Services on linked data
Services on linked data
LD4L Workshop Breakout Session, Tuesday, February 24
facilitator: Jon Corson-Rikert
Risk of not knowing what to search for
- Providing discovery endpoints
-
- =E2=80=98hardened=E2=80=99 SPARQL endpoints may be less prone to down t=
ime =E2=80=93 e.g., Fuseki documenta=
tion states that "authentication and control of the number of concurren=
t requests can be added using an Apache server"
- publishing starting points with examples and standard extracts may help=
=20
- emulate Social Explorer http://socialexplorer.com as a way to q=
uery the contents of a larger data source, in that case census data
- the linked data fragments technology (http://linkeddatafragm=
ents.org) may facilitate hosting linked data without the server-side ov=
erhead and risk of a public SPARQL endpoint
- VIVO/Vitro 'rich export' =E2=80=93 augmenting standard linked data resp=
onses with standard queries=20
- e.g., get all a person's publications from a single request rather than=
client having to issue multiple requests
Synchronizing harvested information
- Risk of harvested or aggregated information going out of sync=20
- Resource sync standard addressed the need to =
repeatedly synchronize and update
- Semantic Web crawling leveraging HTML web crawler experience=20
- what's attached
- what has changed
Desire to be able to query on different axes
=
Reconciliation services
- not necessarily centralized or monopolies
- would work best in an iterative mode, with curation and provenance to m=
anage difference of opinion (or evidence)=20
- who's made that assertion =E2=80=93 differentiate librarians from crowd=
sourcing
- some way to express variable confidence levels
- incorporate feedback from users
- need protocols =E2=80=93 could leverage a common API for reconciliation=
building on the OpenRefine API =E2=80=94 specify as much metadata as =
you have, get ranked results back
- surface (publish) the results =E2=80=93 known servers, as with annotati=
ons =E2=80=93 select which servers to request responses or harvest dat=
a from=20
- notifications of new matches?
- ability to +1 or thumbs-up the connection to corroborate =E2=80=93 Red=
dit gets a lot of traction that way
- repeating assertions in multiple repositories
- =
sameAs.org but with other expressions for and levels of confidence in t=
he relationship
Validation=
h3>
- RDF data shapes working group
- DCMI tutorial on RDF val=
idation
- Measure the consistency of ontology use
- Linked data needs mashup tools that test connections and illustrate bri=
nging data together
Ontology extension mechanisms
=
Ability to push bookmarks
- Small graphs of data, consumable by others, to a platform similar to Me=
ndeley but not limited to bibliographic material
- A service where I can push the results of my search, organized by topic=
- Add things to a collection I have
- Similar to an annotation service
- You search, you refine it, you step back =E2=80=94 now only s=
ave as bookmarks at one level
- Nobody can use your web bookmarks now
- Hide the URIs behind a UI
Additio=
nal ideas
- Semantic autotagging
- Nanopublications =E2=80=93 breaking academic articles int=
o independent assertions with a mechanism to agree/disagree
- Side wikis =E2=80=93 a plugin for the Netscape browser where a wiki cou=
ld be associated with any web page and display additional, user-entered con=
tent or commentary on any web page
- individual libraries will become the authorities for special collection=
s =E2=80=94 items, people, events=20
- queries to a central area would find a match
-
cache the sameAs so don=E2=80=99t have to re-query; everybody who cons=
umes has the cross-links
-
the sort of thing that OCLC might end up doing =E2=80=94 could be any type =
of object =E2=80=94 logical to start with works
- regular expressions to apply against EAD to suggest what is linked to; =
feed into a system to validate, then give pointers to the link
- a clustering algorithm to track the number of times a link between two =
entities is traversed, effectively shortening the distance between them
- a better page rank algorithm for linked data
- anybody a favorite semantic search engine (no =E2=80=93 too siloed)
- visualizations have to be crafted individually
------=_Part_30333_967017012.1711713248053--