35
Crowdsourcing Descriptions for Nature Recordings Maarten Brinkerink Netherlands Institute for Sound and Vision November 22nd – MCN2014 Dallas

Crowdsourcing Descriptions for Nature Recordings

Embed Size (px)

DESCRIPTION

Maarten Brinkerink and Johan Oomen (Netherlands Institute for Sound and Vision, NL) will talk about Waisda?, an open source video labeling game framework developed by Sound and Vision[3], which is currently being developed further in the context of Europeana.[4] Sound and Vision has collaborated with several public broadcasters in the Netherlands to enable fans of certain programmes to contribute fine-grained description of this content. In the latest edition called ‘Spotvogel’ (Mockingbird) Sound and Vision collaborated with the nature TV programme ‘Vroege Vogels’ (Early Birds, by the VARA) to mobilize the online community around the programme for identifying flora, fauna and locations within specific segments of the broadcasts. To support the tagging of the flora and fauna the game utilized a controlled vocabulary that is maintained by Naturalis. Players are awarded points when their tag entries match with other players, and they can score bonus points for using ‘professional’ terms from the controlled vocabulary. Players can also earn badges for certain achievements within the game, for instance for identifying a certain number of birds. Up until now the game managed to gather over 240,000 tags.

Citation preview

Page 1: Crowdsourcing Descriptions for Nature Recordings

Crowdsourcing Descriptions for Nature Recordings

Maarten Brinkerink Netherlands Institute for Sound and Vision

November 22nd – MCN2014 Dallas

Page 2: Crowdsourcing Descriptions for Nature Recordings

Netherlands Institute for Sound and Vision

Page 3: Crowdsourcing Descriptions for Nature Recordings

Our Mission

“As guardian of Dutch audiovisual heritage, we keep Dutch history, as documented in moving images, alive. We enable everyone to utilize the collections to learn, experience and create.”

2014

Page 4: Crowdsourcing Descriptions for Nature Recordings

R&D at Sound and Vision

Page 5: Crowdsourcing Descriptions for Nature Recordings

Audiovisual broadcasts

Professional annotation

Search engine

Television makers

General public

Academics

Page 6: Crowdsourcing Descriptions for Nature Recordings

Television makers

General public

AcademicsProfessional annotation

Audiovisual broadcasts

Search engine

Machine analysis Data gathering

Page 7: Crowdsourcing Descriptions for Nature Recordings

Europeana (Awareness)

Europeana is the trusted source of cultural heritage. Explore millions of items from a range of Europe's leading galleries, libraries, archives and museums. Books and manuscripts, photos and paintings, television and film, sculpture and crafts, diaries and maps, sheet music and recordings, they’re all here.

Europeana Awareness is a Best Practice Network, led by the Europeana Foundation, designed to - among other things - promote its use by a broad public for a variety of purposes including recreation and hobbies, research, learning, genealogy and tourism – engaging users via user generation of content, creation of digital stories and social networking.

Page 8: Crowdsourcing Descriptions for Nature Recordings

Three core WP2 objectives

Research in end-user involvement that will help define opportunities and challenges for Europeana

Launch a two thematic campaigns that each cover a specific challenge for gathering and linking UGC to Europeana

Establish close collaborations with the Wikipedia Community

Page 9: Crowdsourcing Descriptions for Nature Recordings

WP2 – End-user Engagement

“This WP implements support for the meaningful inclusion of User Contributed Content (UCC) content in Europeana and of the distribution of Europeana content in external environments.”

[1] Contextualisation – users adding context to heritage objects in the form of stories and descriptions;

[2] Contribution – gather digital objects from end-users that can help to enrich and compliment the collection on Europeana;

Page 10: Crowdsourcing Descriptions for Nature Recordings

Task 2.1 Tools used to enable end user contributions to Europeana content

Oxford UniversityUsed to contribute stories in the context of 1914-1918

We Are What we Do and PSNC,Used to upload and publish content for 1989

Spild af Tid, NTUADigital Storytelling Platform

Existing tools (Waisda)

Page 11: Crowdsourcing Descriptions for Nature Recordings

Digital Storytelling Platform: Editing a Story

Page 12: Crowdsourcing Descriptions for Nature Recordings
Page 13: Crowdsourcing Descriptions for Nature Recordings

Historypinhttp://www.europeana1989.eu

Page 14: Crowdsourcing Descriptions for Nature Recordings
Page 15: Crowdsourcing Descriptions for Nature Recordings

Wikipedia Edit a thons, 10 countries

Sweden (WW1) – November 7, 2012

Sweden (Fashion) – March 22, 2013

Poland (1989) – June 9, 2013

Denmark (1894) – June 8, 2013

Netherlands, Greece, Australia, Belgium, Germany, Serbia, Sweden and UK (WW1 Edit-a-thons) – June 29, 2013

Sweden (Fashion) – November 12, 2013

Europeana Fashion Editathon at Nordiska museet in Stockholmhttps://commons.wikimedia.org/wiki/File:Europeana_Fashion_Editathon_2013_11.jpg

Page 16: Crowdsourcing Descriptions for Nature Recordings

Wiki Loves Public Art photo competition

• Executed in May 2013

• Sweden, Spain, Austria, Finland and Israel joined the contest in 2013

• 9,250 images were uploaded as part of the contest by 225 uploaders, of which 57 percent were first time contributor

• The articles with photos from the contest have been shown a total of 1,353,909 times between May-October 2013.

Page 17: Crowdsourcing Descriptions for Nature Recordings

Classification of Crowdsourcing Projects

Europeana Awareness: D2.1 User requirements and IPR implications for User Contributed Content in Europeana

Johan Oomen & Lora Aroyo http://www.iisi.de/fileadmin/IISI/upload/2011/p138_oomen.pdf

Correction and

Transcription

ContextualisationCo-curation

Classification and Tagging

Collection acquisition

Page 18: Crowdsourcing Descriptions for Nature Recordings

Classification of Crowdsourcing Projects

Europeana Awareness: D2.1 User requirements and IPR implications for User Contributed Content in Europeana

Johan Oomen & Lora Aroyo http://www.iisi.de/fileadmin/IISI/upload/2011/p138_oomen.pdf

Correction and

Transcription

ContextualisationCo-curation

Classification and Tagging

Collection acquisition

Page 19: Crowdsourcing Descriptions for Nature Recordings

Video Labeling Game – What’s That? (Waisda?)

Allows internet users to annotate audiovisual archive material in the form of a (serious) game

The goal of the game is consensus between players

Fun and competition as motivation

Page 20: Crowdsourcing Descriptions for Nature Recordings

Why?

Investigate the added value of social tagging

Experimenting with new forms of services for the public (serious games)

Which results in:

• Time-related metadata

• Social tags (bridging the semantic gap)

• Interaction between the archive/broadcaster and the public

Page 21: Crowdsourcing Descriptions for Nature Recordings

Spotvogel (‘Mockingbird’) The Third Installment of Waisda?

Based on the Vroege Vogels (‘Early Birds’) nature series by NL public broadcaster VARA (also a partner in the project)

Collaboration with Naturalis, utilizing their Dutch Species Catalogue for matching the social tags to an authoritative taxonomy

Targeted the online community of interest associated with the series (thousands of active online forum contributors, on the programme website)

Page 22: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Homepage

Page 23: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Tagging and scoring

Page 24: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Tagging and scoring (zoomed in)

Page 25: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Tagging and scoring (zoomed in)

Match with user

Page 26: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Tagging and scoring (zoomed in)

Match with user

Vocabulary match

Page 27: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Tagging and scoring (zoomed in)

Match with user

Vocabulary match

Potential match

Page 28: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

Game recap

Page 29: Crowdsourcing Descriptions for Nature Recordings

Game Mechanics

User profile

Page 30: Crowdsourcing Descriptions for Nature Recordings

Results

Three implementations resulted in over a million social tags, by thousands of players

On average 50% of the social tags consists of matched tags, and 25% corresponds to controlled vocabularies

On average 10-20% of the social tags are unique

‘Super taggers’ are responsible for the vast majority of the social tags that are added

Page 31: Crowdsourcing Descriptions for Nature Recordings

Results

The extent to which expert cataloguers deem the social tags to be useful, heavily depends on the type of content

The balance between social tags that correspond with terms from a controlled-vocabulary and terms invented by users themselves, also depends heavily on the type of content

First experiments suggest that the social tags enable high recall fragment retrieval

Page 32: Crowdsourcing Descriptions for Nature Recordings

Lessons Learned

Don’t try to reach a broad audience, but find an active niche

Open knowledge structures provide a way to structure the data that is gathered, and – at the same time – provide great possibilities for linking collections

Crowdsourcing means accepting and respecting multiple authorities and perspective in regards to your collection

Page 33: Crowdsourcing Descriptions for Nature Recordings

Related: eCreative – Sound Connections

-Enrich sounds with Europeana materials and other websources-Invite communities to interact

-http://www.historypin.com/en/explore/birdlife/

Page 34: Crowdsourcing Descriptions for Nature Recordings

Related: eSounds – Wikipedia Editathon

-Enrich Wikipedia with bird recordings-Contextualize sound recordings in a relevant knowledge environment-Bring together Wikipedians & birders

Page 35: Crowdsourcing Descriptions for Nature Recordings

Thanks for your attention!

Maarten Brinkerink

Netherlands Institute for Sound and Vision

[email protected]

@mbrinkerink

@benglabs

http://labs.europeana.eu/apps/Waisda/

https://github.com/beeldengeluid/waisda

Many thanks to:

Johan Oomen & Lizzy Komen, for their input

Just Vervaart, Cyril Snijders, Sander Pieterse, Michiel Hildebrand & Martijn van Steenbergen, for their involvement in ‘Spotvogel’

Let’s continue to think big, y’all!!!