Download presentation
Presentation is loading. Please wait.
Published byColin Gray Modified over 9 years ago
1
Key issues in publishing and consuming linked data for libraries Gordon Dunsire Presented to CILIP Linked Data Executive Briefing 24 November 2015, London
2
Overview Linked data 101 Linked data vocabularies Local vs global Eating cake
3
Item is Person by purchased owned autographed borrowed donated photocopied conserved Linked datalinks 2 specific things Of interest to Libraries Archives Museums Of interest to Publishers Booksellers The Crowd The Cloud Of interest to Semantic Web
4
Person is Item of purchaser owner autographer borrower donor photocopier conserver Linked datagoes both ways Entity relationship
5
Person is Item of purchaser owner autographer borrower donor photocopier conserver Linked dataneeds identities Which person?Which item? “Jane Smith, the author of ‘Article’, OUP, 2001” “The one in the Reference section” “Smith, Jane, 1975-”“0123-456-789” Uniform Resource Identifier {URI} For humans {URI} For machines Global
6
title “Ode to himself” Ben Jonson Place X Parchment This ms author “Jonson, Ben” “abcxyz” birthplace normalised name coordinates material “Requires...” location treatment
7
3 types of linked data vocabulary Datasets Individual things E.g. specific Person, Item, Place, etc. Value vocabularies Concepts, terminologies E.g. subject headings, thesauri, etc. Element sets Types of thing (classes); types of relationship between things E.g. Person, place of birth, supervisor, etc.
8
Linked data vocabularies Each thing is globally identified by a URI A thing may be identified by more than one URI A URI must identify only one thing Each thing is linked to, and humanly identified by, a label and/or definition. A thing may be identified by more than one label A label may identify more than one thing Labels are fashionable, and at the mercy of convention and trend
9
Change and persistent chaos All linked data persists forever Nothing is forgotten Nothing is deleted (but statements can be deprecated) Every statement is copied Change should be well-audited to minimize chaos Every statement is linked to another statement There is no truth out there
10
Who maintains the identifiers (URIs)? LocalGlobal Unique things in datasets Common things in datasets Local value vocabularies External value vocabularies Local element sets Global element sets Persistence requires commitment Global requires availability Trust requires provenance Linked Open Data
11
Closed and open data Closed applications (e.g. local database) Open applications (e.g. Semantic web) URIs not required (blank nodes ok) Permanent sets of triples (aka records) What is not recorded does not exist All things must have a URI (blank nodes not ok) Triples stand on their own What is not recorded has not been recorded yet
12
Unconstrained versions Map of “Audience” umarc: m “adult, general” “adult, serious” pbcore: adult “adult” m21: e “adult” MPAA: NC-17? BBFC: 18? Element sets (schema) Value vocabularies (KOS) Broader/narrower/same? m21: “Target audience of …” m21: “Target audience” frbrer: “has intended audience” schema: “audience” dct: “audience” rdau: “Intended audience” isbd: “has note on use or audience” isbdu: “has note on use or audience” rdaw: “Intended audience” rdfs:subPropertyOf umarc: k
13
Having your cake and eating it Think globally, act locally No global element or value that matches your data? Avoid dumb-down! Publish your own element or value Use open tools Develop and publish maps from your element or value to the nearest global-but-dumber one Maintain your local things for persistent global use (act professionally) Publish your local datasets with local elements and values in a global framework with due diligence
14
Paradigm shifts? From record to data statement From production to consumption From consumer to publisher From closed to open From local to global From smart to dumb From certainty to chaos
15
Legacies Legacy data can be published as linked data without loss of information Legacy elements and values can be mapped to more general vocabularies to interoperate at lowest common semantic level Legacy data cannot be smartened-up There is more future legacy than past New elements and values can be at whatever level of semantic granularity is required
16
Final thoughts We are all in this together At global level, we are an endangered minority Our data is valuable to some industries E.g. advertising, tourism, infotainment We have global infrastructures and expertise for sharing metadata The global linked data Search tool is being invented right now (in a garage basement)
17
Thank you! gordon@gordondunsire.com gordon@gordondunsire.com
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.