svenska yle metadata and data first
TRANSCRIPT
Data first and linked data at the Swedish speaking Yle
Mikael Hindsberg, concept developer svenska.yle.fi @mickhinds | [email protected]
27.5.2015Background image: CC BY-SA http://commons.wikimedia.org/wiki/User:Mschel
Linked data
We now link content at Svenska.yle.fi over:
• Organizational borders
• Content managment systems
• Different languages
• Different media types (text/video/audio)
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
Linked data
WHY?
and
HOW?
… have we done this
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
"I Wonder". CC BY - http://commons.wikimedia.org/wiki/File:I_Wonder.jpg#mediaviewer/File:I_Wonder.jpg
Organization
Svenska Yle is a miniature of Yle
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
• We are a small agile unit who can pilot many things within the company
Organization
One tv-channel and two radio channels- And the web
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
Big disruption in digitalizing media
Media convergence
The audience is changing the way they consume media – FAST!
How to handel this?
Unified platform, Drupal7, 2012
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
News, current affairs, sports, entertainment, lifestyle, recipes, health, science, debate – all in one hierarchically flat platform.
MUST ALSO BE UNIFORM AS DATA!
Referer trafic
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
- Machine readability- Semantic metadata - Linked data- Open data- SEO
The content must be able to live independently from the platform!
Content – i.e. articles – rule!
60% of the trafic starts from an article, 29% from the main page. 11% covers everything else!
Mobile over 50%
Week Desktop Mobile Tablet21/2014 57 % 27 % 16 %21/2015 46 % 37 % 17%
Metadata is the key
• Content, platform and distribution are exploding in diversity.
• We need to be able to serve the web with our content as data.
• Semantically rich ontologies with public URI’s• We use Finto (Finnish ontology and thesauri service) and
Freebase >> Wikidata ca. Aug. 2015• Linked data
• Map relation’s between content to link them together • Data graph
• This graph gives structure, recommendations, search engine optimization, new knowledge, global intercompability
= Data first
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
The components• Finnish ontology and thesauri library
(FINTO) www.finto.fi • Frebase www.freebase.com • Drupal module for annotation:
https://www.drupal.org/project/yild
• + journalists do the base annotation• complemented with automatic annotation
• New module can utilize wikipedia, geonames > almost all open metadata repositories: YILD – Yle Integrator for Linked Data https://www.drupal.org/project/yild
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
YILDYle Integrator for Linked Datahttps://www.drupal.org/project/yild + also check out PoolParty-extension https://www.drupal.org/project/yild_poolparty
Yle-API• On top of our metadata sits an
Yle API-layer (API = application programming interface)
• Meta-API ties together all Yle-metadata to a graph
• API-calls in JSON(-LD)• Compatible with Schema.org, EBUCore och Dublin
Coren
• http://developer.yle.fi/tutorials.html
CC BY-SA http://commons.wikimedia.org/wiki/User:Pbroks13
Roadmap for Yle API
SYND/FYND DRUPAL 7
Programs API
Articles API
Meta API
Image API
IMS
Arena API
Login API
Weather API
NewsGuard
ProgramGuide
Finto
FreebaseScores
APIArpa
Metrics API
The opening of Yle-APIWe have taken the first steps to start opening up our API’s by publishing the Programmes API.
When the Articles- and meta-API’s are opened 3rd party developers can build own versions of most of our services
• http://developer.yle.fi/tutorials.html
Annotation – journalists vs. algorithms
Journalists:
+ abstraction+ logic
- inconsistancy in both quality and quantity- poor attention to detail
Algortihms:
+ attention to detail+ consistancy
- lack of human logic- lack of languag knowledge - idioms- great sense of detail
Uutisvahti – mobile applikationNews application for pushing news stories through metadata >> http://yle.fi/uutisvahti/
Linked data
We can now link content (data) over borders of:
• Organization
• Publishing systems (CMS)
• Language (language neutral, works with Finnish, Swedish and English)
• Media type (text, video, audio, images)
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
Example
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
Linking over language
Termsidor
Swedis
h
Finnish
Example Recomendations
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
• Based on semantic tags• Easy to add more
attributes like metrics• Can be improved by
algorithms, like cos-similarity
• Must be careful to not make too exact recommendations > boring
• Serendipitet, show the audience what they didn’t know they wanted to know
RecomendationsExample
Linking different media types new information from the graph
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
Example
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel
How?It’s really quite simple Below note to developer who built it.
ExampleMedia recommendation, when the article was new:
ExampleThe same content after 3 months. The graph lives!
Hur?Still demands quite complicted api-calls must optimize
Thank you! Questions?
Mikael ’Micke’ Hindsberg
twitter.com/mickhinds
svenska.yle.fi utveckling.ylebloggen.fi
www.slideshare.net/mickhinds
Background image: CC BY SA http://commons.wikimedia.org/wiki/User:Mschel