Art Art History Artificial Intelligence Crypto Philosophy

AI Art, Ownership, Blockchain

Questions of “ownership” in art can be a matter of law, of social norms, or of art theory. New art forms and new methods of producing art can fall foul of existing answers to these questions or creatively re-open them. Often they do both. “AI Art” produced using contemporary “Artificial Intelligence” artificial neural network software is a good example of this. “Rare Art” produced using blockchain token software is another, which we will consider below in relation to one particularly notorious example of AI Art.

“Portrait of Edmond Belamy” was produced by the artist group “Obvious” using an existing artificial neural network model trained on a corpus of images of classical paintings. Obvious did not credit the author of that network model, or any of the artists whose paintings were included in the corpus. At this point there are already three different layers of questions around ownership.

Firstly, the assembly of an image corpus. Accurate reproductions of paintings that are no longer in copyright should not attract copyright, and in the US at least this is quite rightly the case. A collection of such reproductions may attract copyright on the collection itself, but this should not affect individual works within the collection. If the images were of paintings that are still under copyright, copying each image might infringe that copyright. I say “might” because doing so might fall under fair use/fair dealing (hereafter just “fair use”) exceptions to copyright. These exceptions are popular both with artists who work with appropriation and with large Internet companies who work with search and advertising. Both groups, and others such as Digital Humanities scholars, might wish to assemble such corpora of images so it is difficult to generalise about motives and outcomes regarding them. In the case of art, however, fair use for artists is a key defence of artistic creativity in an age where the visual environment is dominated by corporate media.

Beyond this legal view is the ethical and art theoretic view. Is it right to treat individual artworks in a corpus as tokens of a type or as just part of a set, as fodder or as raw material for an industrial process? With apologies to Clement Greenberg, does discarding the tactile elements of painting still meaningfully capture it, and does discarding visual detail and differences in scale discard more for processes of derivation than processes of study?

Secondly, the training and use of the artificial neural network model on that image corpus. The model is trained by processing images in the corpus, by copying and reading their data. The model will contain representations of parts of the images from its training corpus, and its output will also resemble parts of the images they are trained on. Mechanical copying and creating derivatives of images are covered by copyright. Cutting up images and juxtaposing them with the work of other artists is covered by the moral rights that accompany copyright. Again, copyright does not apply to works that are out of copyright (moral rights vary by country…), and artists should have a claim to fair use of such materials. The degree to which artistic use of source materials transforms them should be a factor in establishing that such use is indeed fair use, and the output of artificial neural networks certainly transforms the images that they are trained on. Style is not copyrightable (let’s not talk about “trade dress” here), but forgery and “passing off” can be legal matters, and the application an artist’s signature style to a work that they have not made but is sold under there name is the same whether performed by human hand or algorithm.

Again, the ethical and art theoretic view raises more questions. Signature styles are a matter of pride as well as profit for artists, and while this can be critiqued within art theory it is a strongly established norm that simple imitation of style, without a critical framework for doing so, is a breach of artistic norms. Artificial neural network models need not operationalise an individual artist’s signature style in order to devalue the concept of signature styles in general.

Thirdly, Obvious’s use of existing neural network software to generate an image has caused widespread debate. “Signing” the image with the algorithm used by the artificial neural network software to produce it functions as a double-bluff whatever the intention behind doing so. We know that Obvious produced and sold the image, their attribution is not threatened by this. But it erases the work of both Robbi Barret in producing the model of art that the image is simply a product of and of the artists that the neural network’s model already erases the authorship of, both in terms of attribution and in terms of control of their work (even if from beyond the grave in the case of the corpus artists). Software authors should not be able to control uses of the tools that they produce – Microsoft should not be able to censor your writing using Word or claim joint authorship of everything you write using it. But a trained artificial neural network model is a more complex thing than a text editor from this point of view – it is as much content as it is tool. Microsoft should not be able to tell you how you change the contents of an empty text document, but changing a novel or a painting whether represented physically or digitally may infringe on the copyright and moral rights that it may have. Again fair use should be strongly considered for artistically transformative use of artificial neural network models.

Art theroretically, such direct and uncredited use of existing materials, even materials created by another artist, may count as appropriation art, which is an established category within the arts. Appropriation art is deliberately transgressive, often for critical effect. Appropriating non-art or low-art materials is very different from appropriating canonical art or the art of leading contemporaries but both can be critical moves. Artistic labour can be appropriated directly, in the case of contemporary artists who use studio assistants to produce art under their own signature such as Jeff Koons or Damien Hirst. The signature that the products of this labour are exhibited and sold under is a key part of its erasure. And where software used to make art is free software(/open source), attribution may or may not be a strong social norm but past a certain point that attribution is useful information to have for artistic, critical, and art historical engagement with the work.

Prior to this, Obvious had already encountered the question of ownership and found an answer that led directly to “Portrait of Edmond Belamy” being sold at auction. That answer was based on AI’s twin in contemporary technological hype, the blockchain.

Christie’s discovered Obvious via their work on Superrare, a blockchain-based “Rare Art” platform. Rare Art is named after the “Rare Pepe” project that developed the techniques of using cryptocurrency and blockchain token technology to record limited edition certificates for digital images. This produces “artificial scarcity” and allows a form of ownership for pieces of digital art art that would otherwise be infinitely reproducible. This use of certificates as ownership proxies for art was pioneered by conceptual art. Compared to a flammable piece of paper with a handwritten signature on, the authenticity of a blockchain transaction secured by a not inconsiderable fraction of the world’s computing power each day only increases over time. It may not be entirely clear what the authenticity is of, but the terms and conditions of Rare Art platforms and the community norms of their users and consumers do produce a vivid image of a novel and very strong concept of ownership.

“True digital ownership” on a blockchain secured by cryptographic keys is seen by its proponents as stronger, more trustworthy, and more absolute than previous conceptions of property. This makes AI art a natural fit for Rare Art because each has needs that the other fulfills: ownership in the case of the products of AI art, strongly perceptible uniqueness but also recognisability as art in the case of Rare Art. Sale at auction also provides this kind of closure for the financial value of art, but new art and in particular digital art faces a bootstrapping problem in which it must establish its value in order to be sold at auction but cannot be sold at auction without first establishing its value. Christie’s saw art by Obvious selling on Superrare and could react to that market signal more quickly and with lower risk than with signals from gallery or online sales of physical goods.

It is a truism of International Art English that art questions things. There are many questions in play in both AI Art and Rare Art. They involve the concept of ownership considered in terms of the law, of social norms, and of art theory. The answers to these questions from within each of these realms individually may be obvious and simple to their practitioners, but between them they may be more at odds than each realises. Negotiating this without closing the door to cultural creativity or opening it to corporate exploitation is a task that is of interest far beyond the artworld.

(I am not a lawyer, etc.)

Art History Books

Essay on Essays on Art & Language

INDEX: INCIDENT IN A MUSEUM VI, 1986, Art & Language

“Essays on Art & Language” (1991, revised 2001), Charles Harrison.

The Conceptual Art artists group “Art & Language” formed at the end of the 1960s. Art historian Charles Harrison was a member of the group from the start of the 1970s. In 1991 Harrison published a book on the history of the group up to that point, relating it to the wider history of art during the collapse of the authority of Modernist art criticism and the rise of Postmodernism. Modernist art criticism for Harrison here is the magisterial formalism of Clement Greenberg and their followers. It is an art criticism that produces psychologized readings of the surfaces of artworks for a genteel “spectator” to consume without remainder. It is, as Harrison puts it, “a theory of consumption masquerading as a theory of production” that masks the historical and technical content involved in that production. In response to this, Harrison argues for a historical materialist art criticism that restores Modernism’s dialectical contrast of values by “erasing the edges” of artworks. Doing this makes tractable to art history those “genetic” materials involved in each artworks’ genesis that Modernist art criticism masked with its decontextualized reading of art as primarily expressive rather than representational.

Against this backdrop of the collapsing authority of Modernist art criticism and the demands of a new critical relationship to art, Harrison presents Art & language as being committed to a critical project of going-on without an overarching theory or (as Harrison puts it) in the ruin of that arch. Going-on was a non-theory. Recognizing the contingency of all artistic representation and treating intellectual property and artistic propriety as oppressive expressions of class society, Art & Language produced work that looked very little like existing art not in order to produce saleable novelty but because that is where the work of art took them. When it is no longer possible to “mean the doing” of dominant discourses and there is still no clear path forward, taking action in the face of this dilemma means committing to unpredictable outcomes. The unpredictability of those outcomes may lead out of art into the wider world, or into hard-won moments of artistic autonomy.

The disinterested spectator presupposed by the discursive content of Modernist art criticism as its historical subject exercised a particular set of competences when stood in front of a painting or a sculpture and would use them to describe what they saw (and felt) in a particular way. Art & Language’s work, and organizational structure, was oriented to undermining the spectator’s understanding of and relation to that work in order to create a disenchanted viewer. Harrison examines many examples of this project of disorientation and their discussion of the physical and representational structure of one of Art & Language’s more complex paintings of the 1980s (for Art & Language turned to painting at the turn of the decade to avoid the trap of a “conceptual art” aesthetic that had been quickly recuperated by the artworld) shows these effects at work very clearly:

“This is the painting.” “This is part of the painting.” “This is a representation of the painting.” This is a representation of part of the painting.” “This is a detail of the painting.” “This is a representation of another painting.” [etc.] – p208

Art that seeks a critical autonomy, Harrison argues, must represent some non-trivial aspect of its historical moment. In painting, this must be animated by changes in figure/ground relationships. Historically significant changes in these relationships, in critical terminology, or in the spectator’s competences, must reflect moments of change in relations within class society. At such moments, Harrison argues, it is possible for technical concerns to become suffused by moral concerns. These changes are not and cannot be merely psychological – transgression is not an option to be taken. Even where artistic autonomy is achieved, alterity will be rapidly reclaimed by the mainstream as the example of 1960s Conceptual Art demonstrates.

The history of Art & Language that Harrison tells is not, by their own admission in the introduction to the revised edition of 2001, one that all of the group’s former members would recognize. A chateau in France now houses a museum dedicated to the world’s largest collection of Art & Language artworks. The historical moment at the end of the Cold War that Harrison was writing this particular collection of essays from is long gone, in terms of politics if not actors. And the Nietszchean “litterateurs” that Harrison criticizes for their LARPing of postmodern irony would have no trouble consuming Art & Language’s a bit aesthetic, a bit political, a bit speaking-of-their-own-manufacture, a bit unusual art-objects in the age of Contemporary Art.

But for a historical materialist project of critical art history these objections are part of that project. The strategies that Harrison follows and those that they identify in the art of Art & Language can be lifted up out of the brief moment of the “end of history” and translated into the current epoch even if their iconographic and discursive points of departure cannot. Harrison starts their discussion of the historical context that they situate Art & Language’s work in by identifying “Modernism in two voices”: one in the stentorian tones of the Modernist critic and the secure beholder of authentically expressive abstract art, the other in the more questioning tones of those who notice that the first voice is saying something considerably less secure and universal than it claims to be. What is at stake in the relationship between these two voices, Harrison argues, is nothing less than the moral content of history.

The existence and production of art, Harrison argues, can be justified by the fact that aesthetic oddness or intensity is uniquely revealing. These oddnesses and intensities can serve to emancipate art by producing critical representations of Modernism (or its contemporary equivalent) in order to present the “meanings of the dominated” using the media and aesthetics of high art. This is a direction that the litterateurs have long since recuperated the posture but not the content of. What oddnesses and intensities are possible in the age of global art fairs and freeports, at what levels of the creation of art? What can the second voice of art say with this now?

Art Art History Crypto Reviews

The Rarest Book

The history of rare digital art doesn’t make sense without Rare Pepes.

Pepe the frog is a cartoon character, originally created by Matt Furie, that turned out to be catnip for Internet meme creators. Some of these memes were formatted as trading cards in order to create humorous simulacral cultural fakes called “Rare Pepes” which were shared on imageboards and then sold on eBay and other marketplaces. In reality, digital images are difficult to make “rare”. They circulate as infinitely copyable files on the Internet. There is a “The Simpsons” meme for this, but that’s not what we’re here to talk about.

The Rare Pepe Blockchain Project took the problem of making rare pepes actually rare seriously and ran with it. It catalogues rare pepe images registered as blockchain-stored metadata in small editions of Bitcoin-based Counterparty XCP tokens. Social media clique exclusivity thereby becomes blockchain artificial scarcity. I talked about the economic and social dynamics of this in “Tokenization And Its Discontents“, but it is worth emphasizing (as Jason Bailey and others have) that one of the outcomes of this was the whole “rare digital art” market. While they do represent a valuable alternative to the economic and social dynamics of the existing artworld, the current rare art tokenization platforms amount to a gentrification of the Rare Pepe Blockchain Project, obscuring that more liminal aspects of their origins and discarding some of their possibilities in the process.

“The Rarest Book” is a physical volume created by Eleanora Brizi and Louis Parker collecting 36 series of Rare Pepes, 1774 in total, along with essays that cover the history of the project and put it in context. It’s a fat paperback edition with a striking green cover, as playful and comprehensive as the work it covers. The Rare Pepe Blockchain Project shows the strength of social and memetic content for building community and value in crypto projects. It would be difficult to produce such a book about most other tokenized art platforms, which tend to lack a unifying theme, iconography, or curatorial approach. If you don’t want to view cartoon frog trading cards as conceptually rich contemporary art (although there is always the MODERNPEPE token on the back of the book in that case), step back and look at the project as a whole. This book is an excellent way of doing that and makes a strong case for the interest, value, and alterity of the project.

So order a copy before it becomes even rarer. There were only 300 to start with. Find out more here:

Art Art History Notebook

Problems to be Isolated, Described and Discussed

A & L developed slowly and untidily around a consensus that there were historical and objective problems which could be isolated and described, and thus discussed. This is what distinguished and distinguishes A & L from other artists or artistic formations. A & L saw these problems as matters to be articulated by work, rather than as professional aspects of their social lives to be adopted only once they had left the studio. Conversation, discussion, and conceptualisation became their primary practice, as art.

– p22, “A Provisional History of Art & Language”, Charles Harrison & Fred Orton.

Art Computing Art History Digital Art History Projects

Contemporary Art Daily Text Analysis

cad-wordcloudContemporary Art Daily (CAD) is a leading contemporary art blog that publishes documentation for selected shows of contemporary art. It was started in 2008 by then art student Forrest Nash, who describes the site as follows:

Contemporary Art Daily is a website that publishes documentation of at least one contemporary art exhibition every day. We have an international purview, and we work hard to get especially high-quality documentation of the shows we publish.

Since 2008 CAD has published the details of more than 1800 shows including descriptive text, images of works included, and lists of artists involved in each show.

Nash describes the criteria used for selecting that documentation as follows:

Our criteria for Contemporary Art Daily is complicated and not perfectly reducible, but I like to say that we are generally trying to balance two motives that sometimes conflict with each other. On the one hand, we do have a kind of journalistic motive: we hope to in some way represent the breadth of what is happening in contemporary art, even when a particular artist is not of personal interest to us. On the other hand, we have a curatorial motive, to advance art we believe in and think is important. I am usually more concerned about making a mistake and failing to see or include something than I am accidentally letting something through the filter that doesn’t belong.


As a curated resource, CAD is not a statistically representative population sample of all available contemporary art shows. Like a museum collection, a survey show or a textbook it is a mediated, value-laden view of the artworld. Its popularity demonstrates the appeal of this particular view to contemporary artworld audiences. Analyzing CAD is therefore a way of gaining an insight into one popular view of the contemporary artworld.

The html code of was downloaded in January 2014 and processed with an R script to extract text and information from each post on the site announcing a show that fits their standard format. This data was then loaded by the R code in this file to generate the report you are now reading. For reasons of practicality and clarity Some analysis has been performed on the entire dataset, some on just the most popular entities (…most frequently occurring values) within it.

The presence or absence of surprises in the data may indicate fidelity or bias in the worldview of either Contemporary Art Daily or of the online contemporary artworld audience in relation to each other. The extent to which this generalizes to the culture or the reality of the wider contemporary artworld is open to question. Comparing CAD to the data of a more general art show resource website would provide evidence for this but is outside the scope of the current study. The reader’s intuition will have to suffice on these matters for now.

You can download an archive of the report here in several formats, the html version is by far the best:

Click here to download

The source code is available here:


Art Art History Projects

Simple Word Frequency in Contemporary Art Daily Press Releases

A simple word frequency count of press releases on Contemporary Art Daily (note split city names):


Art History Art Open Data Projects

Exploring Tate Art Open Data 0

Why visualise the Tate’s collection dataset?

The Tate is the UK’s largest art institution. The free and open release of Tate’s collection data shows just how far open data has come in the last decade, and makes a major resource available for study. This resource allows us to follow two lines of investigation.

The first is into the history of art, using the Tate’s collection as a model of art in general, particularly of British art. The Tate’s collection data describes the form, content, attribution and dates of a sample of art from the past several hundred years. This is a history of art, and as long as we place it in its historical context it can be a useful one.

The second is institutional critique, to analyse the Tate’s collection and contrast it with other collections and with other models of the history of art (verbal, data-based or otherwise). Rather than allowing or controlling for the historical context of the data this makes recovering and examining that context the focus.

It’s possible to succeed or fail at each, and neither requires taking the claims of Museums to represent history or of data to represent reality at face value or in a vacuum. Data visualisation and statistical analysis are ways of dealing with datasets that would take a human reader many years to examine. They are forms of rhetoric, but they are also useful tools.

With suitable modesty of aims and suitable reflection on the historical and political contexts which have given rise to our tools and materials, let us begin…

Art Computing Art History Art Open Data Projects

Exploring Tate Art Open Data 2

This is the second in a series of posts examining Tate’s excellent collection dataset. You can read the first part here.The R and R Markdown code for this series is available at .

As before, let’s get started by loading the data.


Movement Artwork Counts

Next let’s load some code to visualize the number of artworks in the collection categorized as being produced by a particular movement each year.


You can see the code in the Git repository above. It loads the Tate collection data files and declares some functions that we can use to plot movement artwork counts.

We can plot the number of artworks from a given movement, for example the Young British Artists (YBAs):

plotMovementFrequency("Young British Artists (YBA)")

Or we can plot the combined counts for multiple movements, for example those since 1800:


Movements Since 1800
These figures are available as PDFs in the Git repository.

Movement Durations

When did a movement start and end, and how long did it last? We can plot this for movements as defined by the date of production of the artworks labelled as being part of that movement in the Tate collection.


First by movement name:

plotMovementDurations(movement.durations.alpha, movement.order.alpha)

Movements By Name

And then by movement start date:

plotMovementDurations(movement.durations.from, movement.order.from)

Movements By Start Date

These figures are also available as PDFs in the Git repository.

Movement Influences

We can use artists who are in two or more movements as links between movements, constructing a network graph of social connections between movements.
Like the Wikipedia data-based update of Alfred Barr’s handmade diagram for the MoMA Cubism & Abstract Art exhibition of 1936 Collectivizing The Barr Model we can extract a family tree (or Rhizome) of influence between art movements and otherwise use network analysis methods to study the social network of art movements:


Movements Connected By ArtistsAgain, this figure is also available as PDFs in the Git repository.


As you can see some of these graphics work better as posters or large-scale PDFs than as bitmaps. There’s much that could be done with curve fitting and comparison of movement artwork counts. And all the techniques of social network analysis can be applied to the graph of artists and movements.

Next we’ll look at artwork genres, which are not explicitly labelled in the collection dataset.

Art History Art Open Data Free Culture Projects

Exploring Tate Art Open Data 1

This is the first in a series of posts examining Tate's excellent collection dataset available at .

I've processed that dataset using code for Mongo DB and Node.js available at .

The R and R Markdown code for this series is available at .

This document has been produced using Knitr. Text in light grey boxes is R code or the output of that code.

Let's get started by loading the data.


That file reads the comma separated value (csv) files containing information about the Tate's collection and generates some useful extra tables of information. Now we have everything in memory we can start examining the collection data.


What can we find out about artists in general?

summary(artist[c("name", "gender", "dates", "yearOfBirth", "yearOfDeath", "placeOfBirth", 
              name         gender                 dates     
 Bateman, James :   2         : 112   dates not known:  59  
 Doyle, John    :   2   Female: 521   born 1967      :  42  
 Hone, Nathaniel:   2   Male  :2894   born 1936      :  38  
 Peri, Peter    :   2                 born 1930      :  36  
 Stokes, Adrian :   2                 born 1938      :  36  
 Wilson, Richard:   2                 born 1941      :  34  
 (Other)        :3515                 (Other)        :3282  
  yearOfBirth    yearOfDeath                      placeOfBirth 
 Min.   :1497   Min.   :1543                            : 491  
 1st Qu.:1855   1st Qu.:1874   London, United Kingdom   : 446  
 Median :1910   Median :1944   Paris, France            :  57  
 Mean   :1887   Mean   :1920   Edinburgh, United Kingdom:  47  
 3rd Qu.:1941   3rd Qu.:1982   New York, United States  :  43  
 Max.   :2004   Max.   :2013   Glasgow, United Kingdom  :  35  
 NA's   :57     NA's   :1309   (Other)                  :2408  
 London, United Kingdom   : 442  
 Paris, France            :  82  
 New York, United States  :  45  
 Roma, Italia             :  22  
 Edinburgh, United Kingdom:  18  
 (Other)                  : 839  

There are more male than female artists, and the yBA and Pop generations lead the births.

Depending on whether we treat place of birth or place of death as more representative, London and Paris are ahead of New York or Edinburgh.

We can smooth out the birth and death dates by grouping them by decade or century.

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
   1500    1860    1910    1890    1940    2000      57 
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
   1540    1870    1940    1920    1980    2010    1309 
sort(table(artist.birth.decade), decreasing = TRUE)
1940 1930 1960 1920 1970 1900 1950 1910 1880 1890 1860 1870 1840 1780 1800 
 363  285  256  255  222  217  197  186  153  151  136  123   77   72   69 
1850 1820 1830 1980 1790 1810 1760 1770 1740 1750 1730 1700 1720 1710 1630 
  69   67   65   58   57   49   45   44   42   38   31   27   15   13   12 
1680 1640 1660 1600 1580 1590 1610 1650 1690 1620 1990 2000 1500 1530 1540 
  10    9    8    6    5    4    4    4    4    3    3    3    2    2    2 
1550 1560 1670 1570 
   2    2    2    1 

   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
   1500    1900    1900    1890    1900    2000      57 
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
   1500    1900    1900    1920    2000    2000    1309 
sort(table(artist.death.decade), decreasing = TRUE)
2000 1980 1960 1990 1970 1940 2010 1920 1930 1950 1900 1910 1840 1860 1880 
 224  191  172  157  140  131  112  102   92   89   80   69   59   59   54 
1850 1870 1890 1820 1830 1800 1810 1780 1790 1700 1760 1770 1750 1720 1730 
  53   49   49   46   44   42   40   24   23   15   14   12   10    7    7 
1740 1680 1710 1640 1690 1620 1650 1660 1570 1670 1600 1630 1540 
   7    6    6    5    5    4    4    4    3    3    2    2    1 

That's quite a different result from that suggested by the yearly results. Decade-wise, birth percentiles are clustered around the turn of the 20th century, deaths around the second world war. But the largest number of births are in the 1930s/1940s with the 1960s coming in second. The deaths look like they reflect the distribution of births, although it would be useful to confirm this statistically.

The maximim birth being in the 2000s doesn't mean that the Tate is collecting child artists, the birth data also includes the years that artist groups were started.

How well is gender represented in the collection?

table(artist.birth.decade, artist$gender)

artist.birth.decade     Female Male
               1500   1      0    1
               1530   1      0    1
               1540   0      0    2
               1550   0      0    2
               1560   0      0    2
               1570   0      0    1
               1580   0      0    5
               1590   1      0    3
               1600   3      0    3
               1610   1      0    3
               1620   0      0    3
               1630   0      1   11
               1640   0      0    9
               1650   0      0    4
               1660   0      0    8
               1670   1      0    1
               1680   0      0   10
               1690   0      0    4
               1700   4      1   22
               1710   0      0   13
               1720   0      1   14
               1730   0      0   31
               1740   1      1   40
               1750   0      3   35
               1760   0      1   44
               1770   0      1   43
               1780   1      5   66
               1790   1      0   56
               1800  10      0   59
               1810   0      2   47
               1820   0      1   66
               1830   1      6   58
               1840   0      5   72
               1850   0      2   67
               1860   1     10  125
               1870   0     15  108
               1880   4     23  126
               1890   4     18  129
               1900   8     38  171
               1910   3     37  146
               1920   2     33  220
               1930   4     38  243
               1940  12     62  289
               1950   2     40  155
               1960   6     77  173
               1970   8     70  144
               1980   3     21   34
               1990   2      0    1
               2000   2      0    1

table(artist.birth.century, artist$gender)

artist.birth.century      Female Male
                1500    2      0    5
                1600    5      1   44
                1700    6      4  157
                1800   13     24  576
                1900   39    293 1667
                2000   22    190  422

The first, unlabelled, column is for artists whose gender is not currently recorded in the data.

As we saw in the summary, there are more male artists than female artists in the Tate's collection. There is no decade or century in which this trend is reversed. The story is slightly different when we look at artistic movements.


The data for artists includes information on

Error in movements$ : 
  $ operator is invalid for atomic vectors

artists movements. If we looked at the artwork data there might be more, but we'll stick with the artists for now.

summary(artist.movements[c("artist.fc", "artist.gender", "", 
                       artist.fc   artist.gender
 Ben Nicholson OM           :  6         :  5   
 Dame Barbara Hepworth      :  5   Female: 27   
 Gilbert Soest              :  5   Male  :324   
 Joseph Beuys               :  5                
 Sir Peter Lely             :  5                
 British School 17th century:  4                
 (Other)                    :326                
 16th and 17th century : 47    
 18th century          : 27    
 19th century          : 63    
 20th century 1900-1945: 95    
 20th century post-1945:124    

 Performance Art                        : 14  
 Conceptual Art                         : 10  
 Netherlands-trained, working in Britain: 10  
 Constructivism                         :  9  
 Body Art                               :  8  
 British Surrealism                     :  8  
 (Other)                                :297  
 16th and 17th century           18th century           19th century 
                    47                     27                     63 
20th century 1900-1945 20th century post-1945 
                    95                    124 
                         Performance Art 
                          Conceptual Art 
 Netherlands-trained, working in Britain 
                                Body Art 
                      British Surrealism 
                          St Ives School 
                         British War Art 
                       Environmental Art 
                            Later Stuart 
                              Abject art 
                  British Constructivism 
                   British Impressionism 
                                Unit One 
                            Grand Manner 
                             Kinetic Art 
                                Land Art 
                      Aesthetic Movement 
                       Camden Town Group 
                      Conversation Piece 
                            Feminist Art 
                        Geometry of Fear 
                         Return to Order 
                          Seven and Five 
                             British Pop 
              Civil War and Commonwealth 
                           Fancy Picture 
                           Fin de Siècle 
                            London Group 
                    New English Art Club 
             Young British Artists (YBA) 
                            Art Informel 
                             Art Nouveau 
                    Auto-Destructive art 
                          Direct Carving 
                      Euston Road School 
                           Newlyn School 
                           New Sculpture 
                             Optical Art 
                                 Pop Art 
              Post Painterly Abstraction 
              Situationist International 
                  Abstract Expressionism 
                           Arte Nucleare 
                  Artist Placement Group 
       Artists International Association 
                        Der Blaue Reiter 
                                De Stijl 
                            Early Stuart 
English-born, working in the Netherlands 
      French-trained, working in Britain 
                    German Expressionism 
                              Grand Tour 
                       Independent Group 
     Italian-trained, working in Britain 
                        Metaphysical Art 
                    Modern Moral Subject 
                          Modern Realism 
                             Neue Wilden 
                   New British Sculpture 
                          Norwich School 
                        Nouveau Réalisme 
                           Origine group 

The artists included in the most movements are some of the grand elders of British 20th Century art. Being in an art movement doesn't improve gender representation.

The most movements are post-1945. Performance art is more popular than Conceptual art, which is interesting given public discussion of state art funding in the UK. “Netherlands-trained, working in Britain” clearly isn't a movement, as with the birth dates the movement name field doesn't always describe a movement per se.

Let's break down gender by movement.

table(artist.movements$, artist.movements$artist.gender)

                             Female Male
  16th and 17th century    5      0   42
  18th century             0      0   27
  19th century             0      0   63
  20th century 1900-1945   0      9   86
  20th century post-1945   0     18  106
movement.gender <- table(artist.movements$, artist.movements$artist.gender)
movement.gender <- movement.gender[order(movement.gender[, 2], decreasing = TRUE), 
movement.gender[1:20, ]

                                Female Male
  Performance Art             0      5    9
  Feminist Art                0      4    0
  Abject art                  0      3    3
  Abstraction-Création        0      2    5
  Constructivism              0      2    7
  St Ives School              0      2    6
  Body Art                    0      1    7
  Camden Town Group           0      1    3
  Kinetic Art                 0      1    4
  Minimalism                  0      1    4
  Rayonism                    0      1    0
  Seven and Five              0      1    3
  Surrealism                  0      1    6
  Unit One                    0      1    5
  Young British Artists (YBA) 0      1    2
  Abstract Expressionism      0      0    1
  Actionism                   0      0    1
  Aesthetic Movement          0      0    4
  Arte Nucleare               0      0    1
  Art Informel                0      0    2

Representation improves slightly over time. Unsurprisingly, feminist art has more female than male artists represented. Abject art is a tie, and there are more than half as many female performance artists as male ones.


There are

Error in eval(expr, envir, enclos) : object 'artwork.title' not found

artworks in the dataset.

summary(artwork[c("artist", "title", "dateText")])
                            artist                    title      
 Turner, Joseph Mallord William:39389   [title not known]: 3659  
 Jones, George                 : 1046   [blank]          : 3520  
 Moore, Henry, OM, CH          :  623   Blank            : 1995  
 Daniell, William              :  612   [no title]       : 1883  
 Beuys, Joseph                 :  578   Untitled         :  627  
 British (?) School            :  388   Mountains        :  540  
 (Other)                       :26493   (Other)          :56905  
 date not known: 5993  
 1819          : 2908  
 1801          : 1331  
 c.1830–41     : 1194  
 1833          : 1171  
 1831          : 1170  
 (Other)       :55362  
   Min. 1st Qu.  Median    Mean 3rd Qu.    Max.    NA's 
   1540    1820    1830    1870    1950    2010    5397 

JMW Turner has tens of thousands more works in the collection than the next nearest artist. Is this a glitch? No, it's due to the fact that the Tate holds the Turner Bequest of around 30,000 works on paper.

What are artworks titled? Usually Untitled, or simply no title. “Mountains” appears to be the most popular actual title, although if we stemmed or otherwise abstracted and clustered the titles other popular ones might emerge.

The most popular years for artworks are in the early 1800s. This, and possibly the titles, are again attributable to Turner. It would probably be productive to remove Turner's works on paper (or more simply just remove all Turner's works) from the dataset and try again, as his presence is clearly skewing the analysis.

Both artists and artworks have movements. Let's look at how artwork movements differ from artists.

 Min.   :    22  
 1st Qu.:  6050  
 Median : 11496  
 Mean   : 21962  
 3rd Qu.: 21954  
 Max.   :114918  

 [no title]                                              : 674  
 [title not known]                                       : 169  
 Untitled                                                : 116  
 Insertions into Ideological Circuits 2: Banknote Project:  54  
 Walking the Dog                                         :  39  
 Exquisite Corpse                                        :  37  
 (Other)                                                 :5894  
      year                   artwork.medium
 Min.   :1545   Screenprint on paper:1301   Min.   :  8    
 1st Qu.:1920   Oil paint on canvas :1113   1st Qu.:290    
 Median :1963   Lithograph on paper : 527   Median :415    
 Mean   :1936   Etching on paper    : 393   Mean   :327    
 3rd Qu.:1973   Graphite on paper   : 205   3rd Qu.:415    
 Max.   :2009   Bronze              : 113   Max.   :415    
 NA's   :303    (Other)             :3331                  
 16th and 17th century : 177    Min.   :  293   British Pop     : 846  
 18th century          : 469    1st Qu.:  363   Conceptual Art  : 445  
 19th century          :1004    Median :  433   Pre-Raphaelite  : 405  
 20th century 1900-1945:1156    Mean   : 2421   St Ives School  : 400  
 20th century post-1945:4177    3rd Qu.: 1683   School of London: 373  
                                Max.   :18626   Neo-Classicism  : 310  
                                                (Other)         :4204  
                British Pop              Conceptual Art 
                        846                         445 
             Pre-Raphaelite              St Ives School 
                        405                         400 
           School of London              Neo-Classicism 
                        373                         310 
                    Pop Art Young British Artists (YBA) 
                        246                         226 
          Independent Group              Constructivism 
                        178                         147 
            British War Art                  Minimalism 
                        141                         138 
            Victorian/Genre           Neo-Expressionism 
                        125                         111 
            Neo-Romanticism      Abstract Expressionism 
                        107                         102 
                 Surrealism            Geometry of Fear 
                         96                          84 
            Performance Art          British Surrealism 
                         81                          75 
 16th and 17th century           18th century           19th century 
                   177                    469                   1004 
20th century 1900-1945 20th century post-1945 
                  1156                   4177 

Pop and Pre-Raphaelitism gain in popularity, but Conceptualism and Surrealism are still popular.


Each artwork is tagged with descriptions of the subjects that it depicts. Subjects have levels, from general to specific, which I've named the category, subcategory and subject. We can group the subjects of artworks by artists and movements to find out what their characteristic subjects were.

summary(artwork.subjects[c("artwork.title", "artwork.dateText", "", 
    "", "")])
           artwork.title          artwork.dateText  
 [title not known]: 13992   date not known: 29732   nature      :76796  
 [no title]       :  8146   1819          : 12948   places      :60314  
 Untitled         :  2148   1833          :  5865   architecture:57507  
 Mountains        :   899   1801          :  5023   people      :52820  
 Shipping         :   462   1831          :  4817   objects     :22990  
 Walking the Dog  :   412   1840          :  4498   society     :20032  
 (Other)          :316957   (Other)       :280133   (Other)     :52557  
 landscape                    : 32722   hill              :  9737  
 adults                       : 22048   wooded            :  8223  
 townscapes, man-made features: 21272   man               :  8164  
 seascapes and coasts         : 12202   figure            :  8118  
 water: inland                : 11839   townscape, distant:  7916  
 countries and continents     : 11704   England           :  7661  
 (Other)                      :231229   (Other)           :293197  
                 abstraction                 architecture 
                       13304                        57507 
emotions, concepts and ideas                      history 
                       11583                         1948 
                   interiors         leisure and pastimes 
                        2467                         3446 
      literature and fiction                       nature 
                        2977                        76796 
                     objects                       people 
                       22990                        52820 
                      places          religion and belief 
                       60314                         4376 
                     society   symbols & personifications 
                       20032                         6242 
        work and occupations                          
                        6214                           NA 
                          NA                           NA 
                          NA                           NA 
                       landscape                           adults 
                           32722                            22048 
   townscapes, man-made features             seascapes and coasts 
                           21272                            12202 
                   water: inland         countries and continents 
                           11839                            11704 
        UK countries and regions cities, towns, villages (non-UK) 
                           10800                            10160 
            non-representational                 transport: water 
                            9583                             9537 
   actions: postures and motions                      UK counties 
                            9055                             8867 
                        features                         military 
                            8694                             7091 
                formal qualities    UK cities, towns and villages 
                            6934                             6695 
              hill             wooded                man 
              9737               8223               8164 
            figure townscape, distant            England 
              8118               7916               7661 
             river              woman           mountain 
              7549               7303               5932 
            castle             bridge              rocky 
              5298               3769               3759 
             group              coast              Italy 
              3694               3545               3509 
     boat, sailing          townscape             colour 
              3381               3157               2859 
               sea              tower 
              2810               2803 

The summary looks like Turner is skewing the results again. The subjects are mostly English landscape of the early 19th Century. But the categories are led by more non-representional subjects, before the subcategories and subjects return to landscape. People (“adults”, “man”, “woman”) emerge as popular subjects as well, indeed they are the second largest subcategory.

summary(artist.subjects[c("", "", "", 
 David Lucas                                    :1653   nature      :991  
 Jacques Lipchitz                               : 301   places      :551  
 Colin Lanceley                                 : 181   people      :471  
 Bernard Leach                                  : 104   architecture:345  
 Langlands & Bell (Ben Langlands and Nikki Bell):  78   abstraction :275  
 Linder                                         :  65   objects     :256  
 (Other)                                        :1091   (Other)     :584  
 landscape               : 287    figure   : 157  
 adults                  : 250    England  : 144  
 weather                 : 198    wooded   : 138  
 non-representational    : 186    cloud    :  99  
 UK countries and regions: 151    man      :  84  
 animals: mammals        : 145    geometric:  74  
 (Other)                 :2256    (Other)  :2777  
                 abstraction                 architecture 
                         275                          345 
emotions, concepts and ideas                      history 
                         132                           16 
                   interiors         leisure and pastimes 
                          18                           36 
      literature and fiction                       nature 
                          36                          991 
                     objects                       people 
                         256                          471 
                      places          religion and belief 
                         551                           70 
                     society   symbols & personifications 
                         153                           51 
        work and occupations                          
                          72                           NA 
                          NA                           NA 
                          NA                           NA 
                    landscape                        adults 
                          287                           250 
                      weather          non-representational 
                          198                           186 
     UK countries and regions              animals: mammals 
                          151                           145 
                  UK counties townscapes, man-made features 
                          129                           118 
UK cities, towns and villages                 water: inland 
                          118                            99 
             formal qualities     from recognisable sources 
                           92                            89 
         seascapes and coasts actions: postures and motions 
                           66                            62 
             transport: water                   residential 
                           57                            49 
       figure       England        wooded         cloud           man 
          157           144           138            99            84 
    geometric         woman       Suffolk          hill        colour 
           74            62            58            52            47 
          cow         river         horse          rain      ceramics 
           40            38            37            35            32 
monochromatic   River Stour         Essex       sunbeam      farmland 
           29            28            27            27            26 

The results from artist subjects don't differ appreciably from the artwork ones. We wouldn't expect any difference, but some artworks have more than one artist or have none, so this introduces variations.

summary(movement.subjects[c("", "", "artwork.title ", "", 
    "", "")])
Error: undefined columns selected
                 abstraction                 architecture 
                        4977                         2831 
emotions, concepts and ideas                      history 
                        4486                          578 
                   interiors         leisure and pastimes 
                         619                          821 
      literature and fiction                       nature 
                         858                         5634 
                     objects                       people 
                        7516                        11828 
                      places          religion and belief 
                        2635                         1296 
                     society   symbols & personifications 
                        4097                         1843 
        work and occupations                          
                        1568                           NA 
                          NA                           NA 
                          NA                           NA 
                          adults             non-representational 
                            4077                             3651 
                formal qualities    actions: postures and motions 
                            2500                             2138 
   clothing and personal effects                     inscriptions 
                            1602                             1402 
       from recognisable sources                             body 
                            1326                             1170 
                       landscape               universal concepts 
                            1161                             1049 
    emotions and human qualities                   social comment 
                             937                              913 
   townscapes, man-made features                      furnishings 
                             898                              868 
reading, writing, printed matter                         features 
                             820                              735 
          woman             man          figure       geometric 
           1854            1649            1197            1191 
         colour    photographic irregular forms     head / face 
           1111             920             563             531 
       standing         England         sitting          female 
            519             503             497             476 
   printed text            text           group        gestural 
            443             428             411             389 
         wooded       landscape        man-made             sea 
            333             305             276             243 

“Insertions into Ideological Circuits 2: Banknote Project” has multiple json records with multiple movements and topics in each, so it's over-represented here. The subjects are still similar, although with more photography.


What can we conclude from this? The collection is dominated by male British pop artists, more from England than from Scotland or the rest of the UK. The subjects of artworks are what one would expect: landscape, human figures, abstracts. The Turner Bequest skews some of the data, and this should be accounted for or addressed in analysis. A few other artworks also skew some results.

Next we'll look more closely at artistic movements with some data visualizations.

Art Computing Art History Art Open Data Free Software Projects

Exploring Art Data: My _MON3Y AS AN 3RRROR | MON3Y.US Review

Reviewing almost 70 artworks quickly and in depth is a challenge. With _MON3Y AS AN 3RRROR | MON3Y.US, I chose the approach of describing each artwork’s notable features and then pulling out themes and commonalities at the end. Halfway through I realised that by changing each description into a standard format, I could write code to parse the descriptions and analyse them to help me find those themes and commonalities. So I did. The code is in R and it’s available here:

The code loads various modules, parses the file and constructs a corpus and matrix from the words in each review. It then outputs various statistics and graphs regarding them.

First up, which terms do I use most frequently, ten or more times:

 [1] "animated" "bill"     "dollar"   "euro"     "glitched" "image"   
 [7] "mapped"   "show"     "texture"  "video"

The most popular subjects are dollar and Euro bills. Art about them shows something about them. It does so using video, animations (whether video, Flash, or HTML5), images, glitch and texture mapping.

Terms I use five or more times:

 [1] "aesthetic"  "animated"   "art"        "background" "banknotes" 
 [6] "bill"       "collage"    "colour"     "dollar"     "economic"  
[11] "euro"       "flag"       "gif"        "glitched"   "graphic"   
[16] "hundred"    "image"      "loop"       "makes"      "mapped"    
[21] "money"      "note"       "piece"      "rendering"  "show"      
[26] "texture"    "video"      "words"

Flags and words join the subjects, hundred unit notes are the most popular, looped animated GIFs, collages and graphics join the forms and figure/ground relations are there with mention of “background”.

Finally let’s look at words I use three or more times:

 [1] "abstract"    "aesthetic"   "album"       "allow"       "american"   
 [6] "animated"    "apparently"  "application" "art"         "background" 
[11] "banknotes"   "bill"        "black"       "blue"        "changing"   
[16] "classic"     "collage"     "colour"      "composite"   "depicted"   
[21] "direct"      "dollar"      "economic"    "effective"   "euro"       
[26] "facebook"    "flag"        "flickering"  "frame"       "gif"        
[31] "glitched"    "google"      "graphic"     "grid"        "html5"      
[36] "hundred"     "image"       "landscape"   "like"        "link"       
[41] "loop"        "love"        "makes"       "mapped"      "million"    
[46] "money"       "monochrome"  "morphing"    "new"         "note"       
[51] "one"         "page"        "patterns"    "piece"       "pixelart"   
[56] "playing"     "polygons"    "possibly"    "price"       "rendering"  
[61] "screen"      "show"        "signs"       "sites"       "stack"      
[66] "style"       "texture"     "time"        "use"         "video"      
[71] "virtual"     "web"         "white"       "words"       "work"       
[76] "yellow"      "zoomed"

No surprises there, except possibly “love”. The code will confuse “Euro” and “European”, so that’s why the US is mentioned but not Europe. Facebook and Google add corporations to the subjects. Colours are added to the formal properties: yellow, blue, white, black. Landscape joins the subjects. And works play, are direct, are classic, have style, an aesthetic, a price, are new. And I weasel about them with “possibly”.

Next lets look at the associations between words. First some obvious ones.


google           love          1990s            age        ambient 
  0.65           0.59           0.43           0.43           0.43


corrupted     miscoloured         nothing          purest            rows 
     0.75            0.75            0.75            0.75            0.75 
   street            look            much           piece         classic 
     0.75            0.52            0.52            0.48            0.41 


carefully    contract   described        form        sale    specific 
     1.00        1.00        1.00        1.00        1.00        1.00 
  another application       price       piece         art 
     0.70        0.49        0.44        0.43        0.36

The corruption found in association with art here is aesthetic, thanks to glitch art.

The word cloud in the next section has some stand-out words. We can look at their associations as well to follow suggestions from within the data.


bill                         1950s 
0.87                          0.33


vimeo     amateur      batter       beach     clipart   commodity 
 0.40        0.39        0.39        0.39        0.39        0.39


dollar                         1950s 
  0.87                          0.38

Videos are mostly on Vimeo. Dollar and bill occur together so there’s no surprises there.

Word clouds are a good way of quickly visualising word frequency. Here’s one of the words in the reviews:


Using the code from my old posts on Vasari’s Lives and on art bloggers we can find the most similar reviews:

Dominik Podsiadly :  JUST DO IT, Jefta Hoekendijk 

Maximilian Roganov :  Jasper Elings, Jefta Hoekendijk, Keigo Hara, Alfredo Salazar Caro | TMVRTX, Mathieu St-Pierre 

JUST DO IT :  Jefta Hoekendijk, Dominik Podsiadly, Lars Hulst 

Mitch Posada :  Dafna Ganani 

Lorna Mills & Yoshi Sodeoka :  Jennifer Chan 
Jasper Elings :  Maximilian Roganov, Curt Cloninger, Adam Braffman, Δεριζαματζορ Προμπλεμ Ιναυστραλια 

Alfredo Salazar Caro | TMVRTX :  Nick Briz, Maximilian Roganov 

Dafna Ganani :  Mitch Posada 

Jennifer Chan :  Lorna Mills & Yoshi Sodeoka 

Jefta Hoekendijk :  JUST DO IT, Maximilian Roganov, Lars Hulst, Dominik Podsiadly 

Keigo Hara :  Maximilian Roganov, Nick Briz 

Ellectra Radikal :  Lars Hulst 

A Bill Miller :  Mathieu St-Pierre 

Nicolas Sassoon :  Lars Hulst 

Curt Cloninger :  Jasper Elings, Nick Briz 
Δεριζαματζορ Προμπλεμ Ιναυστραλια :  Jasper Elings 
Lars Hulst :  Ellectra Radikal, JUST DO IT, Jefta Hoekendijk, Nicolas Sassoon 

Nick Briz :  Alfredo Salazar Caro | TMVRTX, Keigo Hara, Curt Cloninger 

Adam Braffman :  Jasper Elings 

Rollin Leonard :  Maximilian Roganov 

Mathieu St-Pierre :  A Bill Miller, José Irion Neto, Maximilian Roganov 

José Irion Neto :  Mathieu St-Pierre 

Do those make sense to look at the art?

The clustering code from the same old posts produces different groupings:

Cluster 1 : Robert B. Lisek, Geraldine Juarez 

Cluster 2 : Mitch Posada, Nick Kegeyan, Dafna Ganani, Marco Cadioli, Andrey Keske, Guayayo Coco 

Cluster 3 : Rafaël Rozendaal, Adam Ferriss, Aaron Koblin + Takashi Kawashima, Maximilian Roganov, Fabien Zocco, Jasper Elings, Alfredo Salazar Caro | TMVRTX, Anthony Antonellis, Haydi Roket, Keigo Hara, A Bill Miller, Benjamin Berg, Δεριζαματζορ Προμπλεμ Ιναυστραλια, Nick Briz, Vince Mckelvie, Adam Braffman, Rollin Leonard, Mathieu St-Pierre 

Cluster 4 : Dominik Podsiadly, Thomas Cheneseau 

Cluster 5 : Ciro Múseres 

Cluster 6 : Curt Cloninger 

Cluster 7 : Miron Tee, Jan Robert Leegte, Paul Hertz, Jon Cates, León David Cobo, Kamilia Kard 

Cluster 8 : Nuria Güell, Paolo Cirio, Filipe Matos, Agente Doble | UAFC, JUST DO IT, Gustavo Romano, Tom Galle, Cesar Escudero, Jefta Hoekendijk, Gusti Fink, Ellectra Radikal, Aoto Oouchi, Kim Laughton, Martin Kohout, Marc Stumpel, LaTurbo Avedon, Nicolas Sassoon, Erica Lapadat-Janzen, Milos Rajkovic, Rozita Fogelman, Lars Hulst, Yemima Fink, José Irion Neto 

Cluster 9 : Emilio Vavarella 

Cluster 10 : Dave Greber, Lorna Mills & Yoshi Sodeoka, Jennifer Chan, Frère Reinert, V5MT, Addie Wagenknecht, Systaime, Émilie Brout & Maxime Marion, Georges Jacotey

I chose ten clusters arbitrarily. There’s some overlap looking at the two techniques.

I wanted to try out Topic Modelling on the data but an algorithm for choosing the optimal number of topics simply returned the same number as there are documents. So I tried 8, 12 and 20.

12 gave “nice” results:

     Topic 1    Topic 2       Topic 3    Topic 4      Topic 5        
[1,] "video"    "mapped"      "price"    "bill"       "animated"     
[2,] "bill"     "dollar"      "changing" "dollar"     "architectural"
[3,] "dollar"   "texture"     "image"    "love"       "euro"         
[4,] "direct"   "bill"        "show"     "artist"     "glitched"     
[5,] "facebook" "virtual"     "allow"    "google"     "graphic"      
[6,] "faster"   "polygons"    "also"     "money"      "money"        
[7,] "page"     "constituent" "analysis" "monochrome" "zoomed"       
[8,] "abstract" "exploding"   "another"  "pixelart"   "1990s"        
     Topic 6      Topic 7           Topic 8       Topic 9    Topic 10  
[1,] "graphic"    "labels"          "dollar"      "dollar"   "texture" 
[2,] "abstract"   "landscape"       "glitched"    "euro"     "blank"   
[3,] "aesthetic"  "album"           "bill"        "note"     "blue"    
[4,] "album"      "animated"        "video"       "animated" "classic" 
[5,] "apparently" "appears"         "aesthetic"   "bill"     "economic"
[6,] "banknotes"  "art"             "application" "image"    "essay"   
[7,] "european"   "banknotecollage" "colour"      "loop"     "euro"    
[8,] "flag"       "banknotes"       "economic"    "american" "show"    
     Topic 11     Topic 12  
[1,] "bill"       "art"     
[2,] "dollar"     "bill"    
[3,] "video"      "depicted"
[4,] "background" "dollar"  
[5,] "flag"       "labour"  
[6,] "loop"       "video"   
[7,] "reactive"   "words"   
[8,] "roughly"    "1950s"   

The topics are clearer with more words, these are just the first few for each one. I think this is the closest to what I want in terms of discovering what I have written about, although as I say the choice is arbitrary (or at least aesthetic rather than statistical).

Using more code from the Vasari/bloggers posts, we can plot the associations between words:


Changing the parameters and outputting to PDF creates a more detailed and readable graph. It’s fun and inbetween topic modelling and frequency counts for usefulness.

Finally let’s see how I feel about the art with sentiment analysis:

neutral positive 
     66        3 

I do try to find the positive in artworks but there was one that gave me an immediate and visceral negative reaction in the show (you can spot it if you look hard at the reviews). I’m surprised that there are fewer that count as positive. I “love” one of the pieces. Is it in the positive list?

[1] "Martin Kohout" "Marc Stumpel"  "Ciro Múseres"

It’s not. But one of the ones listed does mention “love”, so I don’t know what’s happened there. Sentiment analysis has improved greatly over the last few years, but apparently not in the library I was using.

If I was going to use these techniques to help review art I’d write longer “bag of word” descriptions for each artwork, with fragments of text and individual words acting almost as tags or streams of consciousness, and I would then use topic modeling and clustering to help pull out themes. I’d prefer to use an algorithm to choose the number of topics, as I feel this is more intellectually defensible, but I like the results enough to use it without. I’m disappointed by the performance of the sentiment analysis library I used, next time I’ll try a different one.

Will there be a next time? Yes, the next time I’m reviewing a group show with more than a few artists. Producing this report has been labour intensive, but I’ve a libary of code now and a better understanding of the issues. And I can automate report construction and revision using Knitr, which would allow me to mix Markdown text and R code without hacing to copy and reformat output.