DC2008 Keynote: access to art museums on-line: a role for social tagging and folksonomy

j trant: social tagging and folksonomy access to art museums on-line: a role for social tagging and folksonomy? J. Trant Archives & Museum Informatics http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy access to collections http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Access to Collections From: J. P. xxxxxx@xxxxxx.com Date: Sat, 26 Aug 2006 11:24:43 -0700 To: timeline@metmuseum.org Subject: Looking for a painting Please help: I have been looking on and off for years for this painting. The painting is of a very well dressed renaissance man standing in a room (a library) in front of him on a table is a large hour glass. The painting has very rich colors. I have talked to a lot of people and they have said they have seen this painting but can't remember its name or the name of the artist. Could you please use your resources to find this painting? Reference Questions c.f. Catechism Project, Janney & Sledge

j trant: social tagging and folksonomy Access to Collections Possibly the desired painting Ian Morrison’s “things we just happen to have in our head” method

j trant: social tagging and folksonomy Access to Collections

j trant: social tagging and folksonomy Access to Collections What “J. P.” knows What a Met curator knows Portrait of a Man, ca. 1520–25 painting Moretto da Brescia (Alessandro Bonvicino) Renaissance (Italian, Brescian, born about 1498, died 1554) Oil on canvas; 34 1/4 x 32 in. (87 x 81.3 cm) standing Rogers Fund, 1928 (28.79) man Provenance: Maffei, Brescia (by 1760, as \"Ritratto d'uomo con very well dressed carta in mano, ed Orologio, di Callisto da Lodi\"); by descent to contessa Beatrice Erizzo Maffei Fenaroli Avogadro, Palazzo library Fenaroli, Brescia (by 1853–at least 1857, as by Moretto); her daughter, contessa Maria Livia Fenaroli Avogadro, later marchesa hourglass Fassati, Brescia (in 1862); her son, marchese Ippolito Fassati, Milan (by 1878–at least 1912); [Elia Volpi, Florence, by 1915–16; table sold to Knoedler]; [Knoedler, New York, 1916–28; sold to MMA] rich colors What they both know ...

j trant: social tagging and folksonomy Access to Collections robe Renaissance scholar scroll time trees from MMA proof of concept

j trant: social tagging and folksonomy Access to Collections bangs robe beard border Renaissance cape scholar contemplation elbow scroll hourglass time landscape trees learned man mountain moustache nobleman painting portrait ring volunteer supplied tags robe might help Renaissance from MMA proof of concept

j trant: social tagging and folksonomy Access to Collections What questions do visitors ask of museums?

j trant: social tagging and folksonomy Access to Collections How do users search on-line collections?

j trant: social tagging and folksonomy Access to Collections shark How do users search on-line collections?

j trant: social tagging and folksonomy Access to Collections shark Text How do users search on-line collections?

j trant: social tagging and folksonomy Access to Collections shark Text ? only one shark How do users search on-line collections?

j trant: social tagging and folksonomy Access to Collections CDWA Describing works of art

j trant: social tagging and folksonomy Access to Collections label text curatorial analysis CDWA Describing works of art

j trant: social tagging and folksonomy Access to Collections ≠sharks! http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Access to Collections mis-match between vocabulary of visitor and the museum

j trant: social tagging and folksonomy Access to Collections http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Access to Collections ...not searching for \"museum\" content; they were searching for \"calligraphy\" content. C. Alexander, et al., Beyond Launch: Museum Videos on YouTube, Museums and the Web 2008 http://www.archimuse.com/mw2008/papers/hart/hart.html http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Access to Collections there’s no such thing as “museum content” discovery is institution agnostic

j trant: social tagging and folksonomy research question Can social tagging and folksonomy improve online access to art collections? http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Research Questions gilded animal bejeweled antler rearing deer Artist woman (Nationality, Dates) Title, Date Medium action Dimensions hunting Ownership Notes museum ... luxury documents dog bow horse object leaping huntress users tag J. Trant, Jan 28, 2008 based on Wyman et al. 2006, Fig.2 a model of tagging http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Research Questions An Experiment Called Steve.Museum

j trant: social tagging and folksonomy Research Questions Users Tag Works of Art

j trant: social tagging and folksonomy Research Questions evening dress balloon sleeves black bracelet cape elegant red Users Tag Works of Art

j trant: social tagging and folksonomy Research Questions The Metropolitan Museum of Art Users Tag Works of Art @ MMA

j trant: social tagging and folksonomy Research Questions The Metropolitan Museum of Art evening dress strap elegant black Users Tag Works of Art @ MMA

j trant: social tagging and folksonomy Research Questions users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 Research Process

j trant: social tagging and folksonomy the experiment and our ﬁndings http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Research Questions: Tagging users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 What was tagged?

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger: Works by Institution [all works] The Metropolitan Museum of Art Boston Museum of Fine Arts 14% 13% Cleveland Museum of Art Tate Modern 7% 3% Skirball Denver Art Museum 1% 9% Indianapolis Museum of Art 13% San Francisco Museum of Modern Art 9% Rubin Los Angeles County Museum 6% Minneapolis Institute of Arts of Art 14% 11% Characterizing Tagging: Works Tagged Same MMA 251 in both sets

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger: Works by Institution [all works] 1,784 works The Metropolitan Museum of Art 14% Boston Museum of Fine Arts 13% Cleveland Museum of Art Tate Modern 7% 3% Skirball Denver Art Museum 11 museums 1% 9% Indianapolis Museum of Art 13% San Francisco Museum of Modern Art 9% Rubin Los Angeles County Museum 6% Minneapolis Institute of Arts of Art 14% 11% Characterizing Tagging: Works Tagged Same MMA 251 in both sets

j trant: social tagging and folksonomy Research Questions: Tagging Characterizing Tagging: Works Tagged

j trant: social tagging and folksonomy Research Questions: Tagging range of media Characterizing Tagging: Works Tagged

j trant: social tagging and folksonomy Research Questions: Tagging users folksonomy who assigns what kinds of terms how many terms? are assigned? terms tag works of art assigned 1 how does users' terms tagging differ? assigned 2 Tags per work # of tags # of taggers # tags per tagger total distinct total total tags distinct tags high low average std. dev. high low average std. dev. all works work 1 work 2 work 3 Study Terms Assigned

j trant: social tagging and folksonomy Research Questions: Tagging users folksonomy who assigns what kinds of terms how many terms? are assigned? terms tag works of art assigned 1 descriptive statistics • tagger velocity how does users' terms assigned 2 • tag variation tagging differ? Tags per work # of tags # of taggers # tags per tagger total distinct total total tags distinct tags high low average std. dev. high low average std. dev. all works work 1 work 2 work 3 Study Terms Assigned

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institution Tagger Single-Institution Tagger # of # of # of # of # of tags # of tags works users works users total distinct total total distinct total 1,784 36,981 11,944 1,621 251 56,399 21,557 690 Characterizing Tagging: Basic Descriptive Statistics

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institution Tagger Single-Institution Tagger # of # of # of # of # of tags # of tags works users works users total distinct total total distinct total 1,784 36,981 11,944 1,621 251 56,399 21,557 690 22 tags / 82 tags / Characterizing Tagging: Basic Descriptive Statistics user user

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger Unregistered Users Registered Users Characterizing Tagging: Compare Registered | Unregistered

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger Unregistered Users Registered Users registered users tag more Characterizing Tagging: Compare Registered | Unregistered

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger: Sessions per Registered User 6% 6% 4 or more sessions 4 or more sessions 6% 3 sessions 3 sessions 6% 20% 2 sessions 20% 2 sessions 68% 1 session 68% 1 session Characterizing Tagging: Sessions per Registered User unregistered users only come once by deﬁnition

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger: Tags per Work, Average by Object Type number of tags per object type divided by number of works of that object type highlighted values for samples of fewer than 10 works 40 35 Average Tags / Work Linear (Average Tags / Work) 33.4 30 25 27.1 27.0 25.1 23.4 23.7 20 22.2 21.6 20.7 20.1 15 16.4 16.0 14.5 13.4 10 5 1.5 0 s rs ts s s eo hs re ts ns a e ry er s ile g k ok i ur lo ec ed in tu el th or in ap tio d xt ct Bo co Vi Pr w bj nt p O M W r la Te ite ul Je o- og er O i ed al Pa l Sc Al di ch at ot n d st ix Au ria an W Ar Ph In M ita d e an um l tli gs U st Co in d an w ra s t D Ar created from e iv tagsPerWork-080828.xls at or ec D Characterizing Tagging: Tags Per Work by Object Type

j trant: social tagging and folksonomy Research Questions: Tagging Multi-Institutional Tagger: Tags per Work, Average by Object Type number of tags per object type divided by number of works of that object type highlighted values for samples of fewer than 10 works 40 35 Average Tags / Work Linear (Average Tags / Work) 33.4 30 25 27.1 27.0 25.1 23.4 23.7 20 22.2 21.6 20.7 20.1 15 16.4 16.0 14.5 13.4 10 some works 5 1.5 0 s rs ts s s eo hs re ts ns a e ry er s ile g k ok i ur lo ec ed in tu el th or in ap tio d xt ct Bo co Vi Pr w bj nt p O M W r la Te ite ul Je o- og er O i are harder al ed Pa l Sc Al di ch at ot n d st ix Au ria an W Ar Ph In M ita d e an um l tli gs U st Co in d an w ra s t D Ar created from e iv tagsPerWork-080828.xls at or ec D Characterizing Tagging: Tags Per Work by Object Type

j trant: social tagging and folksonomy Research Questions: Tagging 16% redundant 84% novel Characterizing Tagging: Vocabulary Stabilization?

j trant: social tagging and folksonomy Research Questions: Tagging 16% redundant 84% novel 36,981 11,944 tags distinct Characterizing Tagging: Vocabulary Stabilization?

j trant: social tagging and folksonomy Are tags new terms? http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Research Questions: Museum Documentation research questions: social tagging and folksonomy analysis users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Museum Documentation: Catalogue Records On-line

j trant: social tagging and folksonomy Research Questions: Museum Documentation Museum Data Definition Req. Example Institution The display version of your museum's name y The Metropolitan Museum of Art Image filename The name of the image file that matches this y aps854.jpg object. Institution An identifying number assigned to the object n 11479 ObjectID in your institution, other than an accession number (e.g. the record number from your collections management system). Title/Object The display for of the work's preferred Title y The Gulf Stream Name or Name. Date The display form of the work's date. y 1899 Maker/Culture The display form of the creator's [or y Winslow Homer, 1836–1910 creators'] name[s] (and dates if desired) Medium The display form of the material, support y oil on canvas and/or techniques used to make the work Dimensions The display form of the work's dimensions y 28 1/8 x 49 1/8 in. (71.4 x 124.8 cm) Object Type The type of work, e.g. painting, sculpture, y painting installation Credit Line The display form of any credit to be shown y Catharine Lorillard Wolfe Collection, Wolfe Fund, with the work of art. 1906 Copyright The Display form of any copyright statement n © Jasper Johns/Licensed by VAGA, New York, NY for the work of art. Accession The accession number assigned to the work y 06.1234 Number of art in your museum Notes curatorial or other notes about the work. n The Gulf Stream was based upon studies made during Homer's two winter trips to the Bahamas in 1884–85 and 1898–99. First exhibited at the Pennsylvania Academy of the Fine Arts in Philadelphia in 1900, the picture was subsequently reworked and \"improved\" by the artist. Early photographs show changes to the sea and to the back of the ship, making the composition more dramatic and vivid. The painting was shown in this state at the Carnegie Institute in Pittsburgh in 1900–1901, and then at M. Knoedler and Co. in New York, where the artist placed on the picture the record-asking price of $4,000. There were problems selling the work because of either its high price or its unpleasant subject matter. Homer may have reworked the painting again in the face of this criticism in order to add the rigger on the horizon that signals hope and rescue from the perils of the sea. Museum Documentation: Catalogue Records On-line

j trant: social tagging and folksonomy Research Questions: Museum Documentation publisher Museum Data Definition Req. Example Institution The display version of your museum's name y The Metropolitan Museum of Art Image filename The name of the image file that matches this y aps854.jpg object. title Institution An identifying number assigned to the object n 11479 ObjectID in your institution, other than an accession number (e.g. the record number from your collections management system). date Title/Object The display for of the work's preferred Title y The Gulf Stream Name or Name. Date The display form of the work's date. y 1899 Maker/Culture The display form of the creator's [or y Winslow Homer, 1836–1910 creator Medium creators'] name[s] (and dates if desired) The display form of the material, support and/or techniques used to make the work y oil on canvas format Dimensions Object Type The display form of the work's dimensions The type of work, e.g. painting, sculpture, installation y y 28 1/8 x 49 1/8 in. (71.4 x 124.8 cm) painting dimensions Credit Line The display form of any credit to be shown y Catharine Lorillard Wolfe Collection, Wolfe Fund, with the work of art. 1906 Copyright The Display form of any copyright statement n © Jasper Johns/Licensed by VAGA, New York, NY for the work of art. Accession The accession number assigned to the work y 06.1234 type Number of art in your museum Notes curatorial or other notes about the work. n The Gulf Stream was based upon studies made during Homer's two winter trips to the Bahamas in 1884–85 and 1898–99. First exhibited at the rights Pennsylvania Academy of the Fine Arts in Philadelphia in 1900, the picture was subsequently reworked and \"improved\" by the artist. Early photographs show changes to the sea and to the back of the ship, making the composition more identiﬁer dramatic and vivid. The painting was shown in this state at the Carnegie Institute in Pittsburgh in 1900–1901, and then at M. Knoedler and Co. in New York, where the artist placed on the picture the record-asking price of $4,000. There were problems description selling the work because of either its high price or its unpleasant subject matter. Homer may have reworked the painting again in the face of this criticism in order to add the rigger on the horizon that signals hope and rescue from the perils of the sea. Museum Documentation: Catalogue Records On-line

j trant: social tagging and folksonomy Research Questions: Museum Documentation Museum New Tags Tags ? Documentation Redundant tags Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation 14% yes 86% unmatched Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation sculpture photograph Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation mask portrait Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Multi-Institutional Tagger: Terms Not Found in Multi-institutional Tagger:Documentation Found in Museum Documentation Museum Terms Not 200 150 Times Assigned 100 50 0 an s gn om si w ic da zo J. Trant created from September 18, 2008 Term termsMuseumDocumentation.xls:musdocNULL-freq Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Multi-Institutional Tagger: Terms Not Found in Multi-institutional Tagger:Documentation Found in Museum Documentation Museum Terms Not 200 150 31,949 terms Times Assigned 100 50 86% not found 0 an s gn om si w ic da zo J. Trant created from September 18, 2008 Term termsMuseumDocumentation.xls:musdocNULL-freq Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Multi-institutional Tagger: Terms Not Found inTerms Not Found in Museum Documentation Multi-Institutional Tagger: Museum Documentation [terms assigned more than 50 times, by occurrence] 200 150 Times Assigned 100 50 0 he ct es w al an om n st e en an on re al af e y te o e n m se fe r n e ric s ul le an re rse g pe r ee n d it en n ld ds ye s ue nd lue co ck e de rk ew ca re pa w in e e rio er u tiv sk t il on ap ge der ee c ia in re er io ra r llo et a ra go hi hi at da u e va a xt fa ou io tr m m ric et rs om om nu at flo br m as b lig rv w ho tt pt tr rt bl te gr st sc ra m w w st w lig te o flo cl po ab w d sc an de la k ac bl J. Trant Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Museum Documentation Multi-institutional Tagger: Terms Not Found inTerms Not Found in Museum Documentation Multi-Institutional Tagger: Museum Documentation [terms assigned more than 50 times, by occurrence] Top 100 Not Matched 200 subject 39 150 genre 27 Times Assigned color 10 100 geo-cultural 10 material 9 50 style/period 5 0 he ct es w al an om n st e en an on re al af e y m se te o e n fe r n e ric s ul le an re rse g pe r ee n d it en n ds ye s ue nd lue ld co ck e de rk ew ca re pa w in e e rio er u tiv sk t il on ap ge der ee c ia in re er io ra r llo et a ra go hi hi at da u e va a xt fa ou io tr m m ric et rs om om nu at flo br m as b lig rv w ho tt pt tr rt bl te gr st sc ra m w w st w lig te o flo cl po ab w d sc an de la k ac bl J. Trant Do user tags differ from museum documentation?

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation Are Other Forms of Museum Documentation More Useful? Genres ✓ gallery label text ✓ docent tour talking points / tour scripts ✓ special exhibition catalogue entries ✓ permanent collection guidebook text ✓ curatorial analysis or notes ✓ scholarly article ✓ web site 'feature' texts ✓ audio tours or podcasts ✓ newspaper/popular article

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation Are Other Forms of Museum Documentation More Useful? Genres ✓ gallery label text ✓ docent tour talking points / tour scripts ✓ special exhibition catalogue entries ✓ permanent collection guidebook text ✓ curatorial analysis or notes ✓ scholarly article ✓ web site 'feature' texts both scholarly ✓ audio tours or podcasts + general public ✓ newspaper/popular article

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation 11 documents 19 distinct tags found 138 times 51 relevant 63% false matches 87 not relevant ... the house where he grew up ...

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation ➡Extended documentation not better than online museum documentation ➡Text mining not algorithmic ➡Public genres not much better than scholarly

j trant: social tagging and folksonomy Do Museum Professionals ﬁnd tags useful? http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Research Questions: Usefulness research questions: social tagging and folksonomy analysis users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 Do Museum Professionals Find Tagging Terminology Useful?

j trant: social tagging and folksonomy Research Questions: Usefulness if you found this work using this term in a query would you be surprised?

j trant: social tagging and folksonomy Research Questions: Usefulness 36,931 terms if you found this work using this term in a query would you be surprised?

j trant: social tagging and folksonomy Research Questions: Usefulness 12% no 88% yes Would Terms Be Useful for Searching?

j trant: social tagging and folksonomy Research Questions: Usefulness In-depth study of a sample Sample Works: Useful terms

j trant: social tagging and folksonomy Research Questions: Usefulness In-depth study of a sample

j trant: social tagging and folksonomy Research Questions: AAT/ULAN research questions: social tagging and folksonomy analysis users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN ✓representational Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN 29.8% unmatched 70.2% matched Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN Multi-institutional Tagger: Terms Found in AAT: Distinct Terms [11,944 distinct terms: 4,448 Matched Some Field in the AAT] 37.2% Matched matched37.2% 62.8% Unmatche unmatched d Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN Multi-institutional Tagger: Full Matches to AAT by Hierarchy Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN Multi-institutional Tagger: Full Matches to AAT by Hierarchy Materials Styles and Associated Periods Concepts cultures Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN 2% full match 34% partial match 64% unmatched Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN ULAN Name: Green Terms are Unambiguous Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Research Questions: AAT/ULAN wood white brown forest winslow homer rembrandt ULAN Name: Green Terms are Unambiguous Are tags in Museum Vocabulary Sources?

j trant: social tagging and folksonomy Are tags useful for searching? http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Research Questions: Searching users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 Could tagging improve on-line searching for works of art?

j trant: social tagging and folksonomy Research Questions: Searching New Tags Tags could Tags Redundant improve search Tags ? ? Unsuccessful Search Museum Museum Documentation Searches Terms Documentation not used in search ? Successful searches Could tagging improve on-line searching for

j trant: social tagging and folksonomy Research Questions: Searching Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching shark Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching Multi-Institutional Tagger: Search Terms in Logs 250,000 Duplicated Search Terms Distinct Search Terms 200,000 150,000 100,000 50,000 0 MIA SFMOMA Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching Multi-Institutional Tagger: Matches to Search Terms in Logs 250,000 Duplicated Search Terms that Match Distinct Search Terms that Match 200,000 150,000 100,000 50,000 0 MIA SFMOMA Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching Multi-Institutional Tagger: Matches to Search Terms in Logs 250,000 Duplicated Search Terms that Match Distinct Search Terms that Match 200,000 150,000 100,000 few tags match 50,000 0 MIA SFMOMA Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching SFMOMA and MIA: Top 100 Search Terms: Compare Frequency by Category 40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% ev st ib nt ge n ti Percent of Searches lt e e io MIA ar l cu nr ra is s it ga ni s b SFMOMA iz ue u c jo ar ion ti g h q pe re in ex su od se at ia log of t o o- ch or ch ec t st ri ar ge n e bj /t e/ k yl ed or st w m Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching SFMOMA and MIA: Top 100 Search Terms: Compare Frequency by Category 40.0% 35.0% 30.0% 25.0% 20.0% mostly artists 15.0% 10.0% 5.0% 0.0% ev st ib nt ge n ti Percent of Searches lt e e io MIA ar l cu nr ra is s it ga ni s b SFMOMA iz ue u c jo ar ion ti g h q pe re in ex su od se at ia log of t o o- ch or ch ec t st ri ar ge n e bj /t e/ k yl ed or st w m Do Tags Match Search Terms?

40.0% 35.0% 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% ar ? ar ti st tis da ge t te n re ge o- d cu ate lt u ra ev l ex en h t ib it io ge n ge n ge n re n st re re yl da e/ t ge ge pe e o o- ri ge -cu cu od o- lt lt cu ur u al ra lt u ge l ra l s nre u bj ec t m jo Searches by Category: MIA ed lo bs ia gi /t st ec ic h s or n j trant: social tagging and folksonomy ga iq n u iz e MIA Top 100 at se io MIA Middle 100 ar n ch in MIA Single Searches g st st yl or e/ e pe ri od Do Tags Match Search Terms? su w bj or ec k t of ar t Research Questions: Searching

j trant: social tagging and folksonomy Research Questions: Searching Searches by Category: MIA 40.0% 35.0% mostly artists MIA Top 100 30.0% MIA Middle 100 MIA Single Searches 25.0% 20.0% 15.0% genre and culture 10.0% 5.0% 0.0% ? ge t re st tis cu ate ev l n ra t ar n en re re d io u pe e cu od lt t n ge l it da ra l s nre ge ib t ri bs ec ti u s o- h ar e lt ic bj jo ex n u ge g n e/ st io e iq u in ge od al or gi o- at t yl n te ch ec t cu ur h lo st ri ar ge iz st ra da ec ar bj pe lt n u re of ge -cu ga /t se su lt e/ n k ia or ge o yl or ed ge o- st w m Do Tags Match Search Terms?

ar 50.0% 45.0% 40.0% 35.0% ti 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% st 0.0% ge o- ar ? ar cu tis ti ltu t da st r te ge al ge nr e o- d da cu at te ltu e su ra bj l ec ex ev t hi en bi t ti ge nr o ge ge n e st nre nre ge yl e d ge /p ate ge o-c o e o- ul -cu rio cu tu lt d lt ral ura ur al gen l su re bj ec t fa ct jo m bs ed ju Searches by Category: SFMOMA ia log nk /t i ec sti j trant: social tagging and folksonomy or h cs ga ni ni qu z e se ati ar on ch in Top 100 st g yl st o Middle 100 e/ pe re ri Single Searches su od w Do Tags Match Search Terms? or bje k c of t ar t Research Questions: Searching

ar 50.0% 45.0% 40.0% 35.0% ti 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% st 0.0% ge o- ar ? ar cu tis ti ltu t da st r te ge al ge nr e o- d da cu at te ltu e su ra bj l ec ex ev t hi en bi t ti ge nr o ge ge n e st nre nre ge yl e d ge /p ate ge o-c o e o- ul -cu rio cu tu lt d lt ral ura ur al gen l su re bj ec t fa ct jo m bs ed ju Searches by Category: SFMOMA ia log nk /t i ec sti j trant: social tagging and folksonomy or h cs ga ni ni qu z e se ati ar on ch mostly artists in Top 100 st g yl st o Middle 100 e/ pe re ri Single Searches su od w Do Tags Match Search Terms? or bje k c of t ar t subject + work Research Questions: Searching

j trant: social tagging and folksonomy Research Questions: Searching Multi-institutional Tagger: Compare Tags to Search Terms: Top 100 Full Matches by Category 45.0% 40.0% MIA 35.0% SFMOMA 30.0% 25.0% 20.0% 15.0% 10.0% 5.0% 0.0% ts r lo ? ti e on ar co al nr s ti ur l ic ge ia bi d st t t io er ul hi ec e gi r t at c ex pe bj da lo o- m su e/ ge yl st Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching Multi-institutional Tagger: Compare Tags to Search Terms: Top 100 Full Matches by Category 45.0% 40.0% MIA 35.0% SFMOMA 30.0% colour 25.0% 20.0% 15.0% genre and subject 10.0% 5.0% 0.0% ts r lo ? ti e on ar co al nr s ti ur l ic ge ia bi d st t t io er ul hi ec e gi r t at c ex pe bj da lo o- m su e/ ge yl st Do Tags Match Search Terms?

j trant: social tagging and folksonomy Research Questions: Searching Could You Search Using Tags?

j trant: social tagging and folksonomy Research Questions: Searching pie painting portrait 20th century white coconut cake background lemon merangue balanced three genoise balancing trio Thiebaud food cherry on top baked goods tall stands frosting simple desserts shadows dessert cake stand cake only 4 tags found this work plates cakes shadow cream in the middle cream ﬁlling lollypop shadows Could You Search Using Tags?

j trant: social tagging and folksonomy Summary 85%+ Tags Are Not Found in Museum Documentation 60%+ Don’t Match Vocabularies Most Tags Can’t Be Mined From Other Sources

j trant: social tagging and folksonomy Can we learn from tagging? user ? generated metadata Understand what users see + what they’re looking for

j trant: social tagging and folksonomy Can we learn from tagging? ? Understand what users see + what they’re looking for

j trant: social tagging and folksonomy Public Vocabulary is Different. Tagging does Contribute. http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Steve Tagger WARNING: extra slides beyond this point http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Can we learn from tagging? Visitor Feedback http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Can we learn from tagging? when given the choice to comment on the exhibition or works with in it, nine out of ten people chose to comment on a speciﬁc piece. Bernstein, S., Where Do We Go From Here? Continuing with Web 2.0 at the Brooklyn Museum, in J. Trant and D. Bearman (eds.). Museums and the Web 2008: Proceedings, Toronto: Archives & Museum Informatics. Published March 5, 2008. Consulted March 20, 2008. http://www.archimuse.com/mw2008/papers/bernstein/bernsteinhtml echoed by Samis, P., Who Has The Responsibility For Saying What We See? Mashing up Museum, Artist, and Visitor Voices, On-site and On-line, in J. Trant and D. Bearman (eds.). Museums and the Web 2008: Proceedings, Toronto: Archives & Museum Informatics. Published March 5, 2008. Consulted March 20, 2008. http://www.archimuse.com/mw2008/papers/samis/ samis.html Point of Intersection http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Can we learn from tagging? Visitor Engagement http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Can we learn from tagging? Visitor Context: steve art tagger in Facebook http://conference.archimuse.com/jtrants

j trant: social tagging and folksonomy Can we learn from tagging? “our catalogues may be out of date before they have left the press”, 1910 http://conference.archimuse.com/jtrants

Research Questions: Tagging Multi-Institution Tagger # tags per tagger # of # of works # of tags total tags distinct tags taggers total distinct total high low average std. dev. median high low average std. dev. median 1784 36981 11944 1621 1903 1 22.81 73.59 6 1127 1 19.09 49.1 6 Characterizing Tagging: Basic Descriptive Statistics Multi-Institution Tagger

Research Questions: Tagging MMA Tagger # tags per tagger # of works # of tags # of total tags distinct tags taggers total distinct total high low average std. dev. median high low average std. dev. median 251 56399 21557 690 1794 1 81.7 172 27 1108 1 64.0 115.3 25 Characterizing Tagging: Basic Descriptive Statistics MMA Tagger

Research Questions: Tagging MMA Tagger: Tags per Work, Average by Object Type number of tags per object type divided by number of works of that object type n.b. No Audio-Video, Installations, Mixed Media, or Other; highlighted values for less than 5 450 393.2 Much more variation in 400 383.0 MMA tag vocabulary Average Tags/Work 350 Paintings different from 300 other types @ MMA 239.8 250 224.7 214.6 209.3 205.5 205.0 200 182.0 174.8 155.8 150 100 50 0 ) 2) 3) 5) 1) 3) ) ) 6) ) ) (1 1 3 (5 (2 (1 (4 (1 5 (9 (1 (4 (2 (2 s ry e ile rs gs s ur ts ts e s el ok ph ur ks lo ec xt in ct tin w pt Bo or co ra Pr Te ite bj Je in ul lW og er O ch Pa d Sc at ot an n Ar Al ria W Ph e a d um lit an tli st gs U Co in d an w ra ts D Ar e iv at or ec J. Trant D created from Sept 4, 2008 tagsPerWork-080828.xls J. Trant created from Characterizing Tagging: Compare MMA Tagger July 16, 2008 tagsPerWork-080714.xls

Research Questions: Tagging Compare Taggers: Tags per Work, Average by Object Type [all works] number of tags per object type divided by number of works of that object type n.b. MMA has no Audio-Video, Installations, Mixed Media, or Other; highlighted values for less than 10 works 393.2 400 383.0 Multi 350 MMA 300 239.8 250 224.7 214.6 205.0 209.3 205.5 200 182.0 174.8 155.8 150 100 50 33.4 27.1 27.0 25.1 23.4 23.7 22.2 21.6 20.7 20.1 16.4 13.4 16.0 14.5 1.5 0 es rs gs s ks o ry hs ns er ia ks e ts re ct e ur ed lo in el il th u or tin o ap tio id e xt ct pt Bo co Pr w bj O V M W gr in la Te ite Je ul o- er O ed al Pa to l Sc Al di ch at d n st o ix an Au ia W Ar Ph In M ar d e lit an m tli tu gs U s Co d in an w ra ts D Ar e iv at J. Trant created from or ec Sept 4, 2008 D tagsPerWork-080828.xls Characterizing Tagging: Tags Per Work by Object Type

Research Questions: Tagging Multi-Institutional Tagger: Works by User coverage [1,784 works 1,621 Total Taggers] @ 4 users 203 200 191 190 185 178 156 Count of total taggers 150 Linear (Count of total taggers) 138 Number of Works 110 109 100 100 61 50 50 28 21 12 13 12 6 6 5 2 2 4 1 0 0 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 22 24 25 28 Total Users J. Trant created from Characterizing Tagging: Users per Work Aug 28, 2008 tagsPerWork-080828.xls

Research Questions: Tagging Characterizing Tagging: Users per Work

Research Questions: Tagging Multi-Institutional Tagger: Works by Tagger by Object Type coverage eleven taggers tagged a painting 24 times | 24 paintings were tagged by 11 taggers each @ 4 users 90 80 70 Architecture 60 Audio-Video Number of Works Tagged Books Costume and Jewelry 50 Decorative Arts and Utlilitarian Objects Drawings and Watercolors Installations 40 Mixed Media Other Paintings Photographs 30 Prints Sculpture Textiles 20 10 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 22 24 25 28 Total Taggers J. Trant created from Aug 28, 2008 tagsPerWork-080828.xls Characterizing Tagging: Taggers per Work by Object Type

Research Questions: Tagging Compare Taggers: Average Number of Tags per User by Object Type [all works] tags per object type divided by number of taggers for that type; highlighted values for less than 10 works n.b. MMA has no Audio-Video, Installations, Mixed Media or Other 5.0 4.7 Threshold may be about tagging, not about works Compare tag vocabulary across object type: are there larger pools of terms for some? 4.5 3.9 4.0 4.0 3.9 3.9 4.0 3.8 3.6 3.6 3.6 3.6 3.9 3.5 3.6 3.6 3.6 3.5 3.5 3.3 3.3 3.2 Number of Tags 3.0 3.2 3.2 Multi 3.0 2.9 MMA 2.5 2.6 2.0 1.5 1.5 1.0 0.5 0.0 es s gs ks ts o ry s er ns ia s e ts e or ph ok de ur ur ed ec in el th til or tin tio ol ct pt Bo ra Vi Pr w x bj O M lW c in la Te ite Je ul o- og er O ed al Pa Sc Al di ch at ot d n st ix an Au ria W Ar Ph In M ita d e an m l tli tu gs U s Co d in an J. Trant created from w ra Object Type ts Sept 4, 2008 tagsPerWork-080828.xls D Ar e iv at or ec D Characterizing Tagging: Tags Per Tagger by Object Type

Research Questions: Tagging To Do: Remove steve team members Multi-Institutional Tagger: Average Tags Per Tagger Per Work, by Object Type and Institution [works tagged by at least 4 taggers] 8.0 Boston Museum of Fine Arts 7.0 Cleveland Museum of Art Denver Art Museum 6.0 Indianapolis Museum of Art Los Angeles County 5.0 Museum of Art Minneapolis Institute of Arts 4.0 Rubin San Francisco Museum of Modern Art 3.0 Skirball Tate Modern 2.0 The Metropolitan Museum of Art Grand Total 1.0 0.0 al s s ts o ia gs s s ns e ry ts e e or ok ph de ur ur t ed ec in til el tin To tio ol ct pt Bo a i Pr w x bj M -V gr c in lla Te te ul Je nd er O io ed Pa o i ta Sc ch at ra ot d n d ix s Au ria an W Ar G Ph In M a d e lit an um tli gs U st Co d in an w ra ts D Ar e iv at or ec D J. Trant created from July 16, 2008 tagsPerWork-080714.xls Characterizing Tagging: Tags per Work by Object Type

Research Questions: Tagging Multi-Institutional Tagger: Average Tags Per Tagger Per Work, by Object Type and Institution [works tagged by at least 4 taggers; n.b. read in conjunction with number of works per type] Architecture Boston Museum of Fine Arts 7 Cleveland Museum of Art Grand Total Audio-Video 6 Denver Art Museum Indianapolis Museum of Art 5 Los Angeles County Museum of Textiles Books Art 4 Minneapolis Institute of Arts Rubin 3 San Francisco Museum of Modern Art 2 Skirball Sculpture Costume and Jewelry 1 Tate Modern The Metropolitan Museum of Art 0 Grand Total Prints Decorative Arts and Utlilitarian Objects Photographs Drawings and Watercolors Paintings Installations Other Measures: • views by object type Mixed Media • skipped by object type J. Trant created from Sept 3, 2008 tagsPerWork-080828.xls Characterizing Tagging: Tags per Work by Object Type

j trant: social tagging and folksonomy Research Questions: Extended Museum Documentation Paul Gauguin French, 1848–1903 Landscape near Arles, 1888 oil on canvas Gift in memory of William Ray Adams 44.1 Detailed review of example

Research Questions: Extended Museum Documentation no. of no. of term no. of no. of terms no. of terms n/a term documents documents frequency documents applicable or string string n/a haystack 3 3 0 0 7 0 tranquil 1 0 0 1 0 3 house 1 1 2 2 2 13 ﬁeld 1 0 0 0 0 5 blue trees 1 0 0 0 0 0 vibrant 1 0 0 1 0 1 color 1 0 2 3 3 20 chemical colors 1 0 0 0 0 0 winters feeding 1 0 0 0 0 0 farm 1 3 1 1 6 10 rural 1 2 0 1 2 3 Fauvism 1 0 0 0 0 0 landscape 1 11 0 3 27 19 ﬁne brush 1 0 0 0 0 0 cypresses 1 2 0 0 2 0 farmstead 1 0 0 0 0 0 blue 1 1 0 2 2 6 orange 1 0 0 1 0 1 yellow 1 0 0 1 0 6 23 5 16 51 87 Summary of Review

Does interface inﬂuence tagging?

Environments research questions: social tagging and folksonomy analysis users folksonomy vocabulary sources who assigns what kinds of terms are terms found in... how many terms? are assigned? Art and Architecture Thesaurus terms tag works of art assigned 1 museum (AAT) records Union List of Artists Names how does users' terms (ULAN) tagging differ? assigned 2 searches of art image databases works of art J. Trant, February 2008 are terms applicable to based on Trant 2006 How does the tagging interface inﬂuence tags assigned?

j trant: social tagging and folksonomy Environments Steve Tagger evening dress balloon sleeves black bracelet cape elegant red Do users tag differently when they don’t see others’ tags or museum documentation?

j trant: social tagging and folksonomy Environments Steve Tagger Do tags vary when museum documentation is shown?

j trant: social tagging and folksonomy Environments Steve Tagger Do users tag differently when they can see what others have done?

Environments Do users tag differently when they see tags and documentation?

Environments Do users ‘get in the groove’ when they tag groups of like works?

Environments works / No. Envir works terms terms terms / works / sess. works reg. sessi sessions Term Set Start End of onme Description works untagge users terms reg. black- work session (avg) tagged users ons reg. users Days nt d users listed (avg) (avg) reg. users Term Set 26- 10- 106 1552 1227 325 1309 220 519 175 10995 5406 4 8.4906 5.1426 6.2795 1 Mar-07 Jul-07 random + no 1 329 1223 320 51 121 37 1730 875 4.845 3.0333 3.5405 metadata random + 2 667 885 319 53 134 38 3902 2065 5.6132 6.8516 9.6316 metadata sets + no 3 526 1026 352 58 130 40 2792 1179 4.9506 5.4724 6.825 metadata sets + 4 578 974 317 66 134 46 2571 1287 4.2837 5.0846 5.2391 metadata Statistics for Term Set 1 BAD DATA

Environments works / No. Envir works session terms terms terms / works / sess. works reg. session Term Set Start End of onme Description works untagge users s reg. terms reg. black- work session (avg) tagged users s Days nt d users users listed (avg) (avg) reg. users Term Set 11- 15- 96 1550 1500 50 1314 333 662 307 17618 11502 4 10.4107 7.7049 11.3478 2 Jul-07 Oct-07 sets + no metadata 2 [continuing 595 955 330 75 162 65 3273 1415 4.9429 4.8086 5.3231 from Term Set 1] sets + no 3 metadata + 553 997 338 86 154 63 2632 1494 4.2893 4.82 6.7302 tags sets + metadata 6 [continuing 1341 209 346 92 196 110 8814 7173 5.7681 13.6872 19.6273 from Term Set 1] sets + 7 metadata + 670 880 324 80 150 61 2899 1420 3.8284 5.9048 7.6066 tags Statistics for Term Set 2 BAD DATA

Environments works / works sessions terms terms terms / works / sess. Term No. of Environ Descrip works reg. Start End works untagge users sessions reg. terms reg. black- work session (avg) Set Days ment tion tagged users d users users listed (avg) (avg) reg. users Term 15- 13- 150 1784 1467 317 2030 322 709 306 13763 7959 12 8.7151 5.1426 6.9498 Set 3 Oct-07 Mar-08 Image hub 9 666 1118 518 78 187 71 4334 2441 6.036 4.7204 6.4366 without tags Image 10 hub 572 1212 491 72 164 68 2582 1546 4.208 4.2981 5.8824 with tags Word hub 11 675 1109 507 81 175 76 3493 2256 4.877 5.4186 7.3553 without tags Word 12 hub 765 1019 535 91 183 84 3240 1834 4.0078 6.1038 7.881 with tags Statistics for Term Set 3 BAD DATA

j trant: social tagging and folksonomy Research Questions: Tagging total tags distinct tags 90.0 67.5 45.0 22.5 0 multi-institution single institution Characterizing Tagging: Basic Descriptive Statistics

j trant: social tagging and folksonomy Environments Steve Tagger average number of tags per tagger no metadata metadata 0 2.5 5.0 7.5 10.0 Term Set 1: Taggers are supplying more tags without metadata (avg 5.75) than with metadata (avg 4.5). Term Set 1: metadata vs. no metadata BAD DATA

j trant: social tagging and folksonomy Environments Steve Tagger average number of tags per tagger no sets sets 0 2.5 5.0 7.5 10.0 Term Set 1: Taggers are supplying far more tags with sets (6317 | 5.8 avg per user) than no sets (3882 | 4.6 per user). Term Set 1: sets vs. random BAD DATA

DC2008 Keynote: access to art museums on-line: a role for social tagging and folksonomy

0 comments

Notes on slide 1

1 Favorite