What is The Art Genome Project? must see for all artists

The Art Genome Project

From Wikipedia, the free encyclopedia
The Art Genome Project
Key people
Carter Cleveland
Sebastian Cwilich
Matthew Israel
Joe Kennedy
Website artsy.net/theartgenomeproject

The Art Genome Project is the search technology behind Artsy.

The Art Genome Project’s search technology is the product of an ongoing art-historical study—undertaken by a team of contributors with art-historical backgrounds at Artsy—seeking to define the characteristics which distinguish and connect works of art, architecture, ancient artifacts and design.[1]

Its primary aim is to provide Artsy users dynamic search categories and explain similarities among art and artists. Currently, there are 1000+ “genes” (i.e. attributes of art) in the project’s taxonomy, including art-historical movements, subject matter, and formal qualities.[2][3] These genes are the product of Artsy’s team and their engagement with (and feedback from) the museums, galleries, curators, critics and art historians present on Artsy’s platform.

There are two general parts of the project: 1) Conceiving and defining such characteristics, referred to as “genes,” and 2) Applying these genes to artists and artworks—creating “genomes” for both—for the artsy.net [4] site.

Importantly, unlike tags, which are binary, genes are applied with values ranging from 0 to 100.[5][6] The value indicates the degree of relevance of a gene to an artist or work of art. While not seen by users, such gene values account for the strength of a relationship between artists and artworks. It also enables similarity to be expressed in a more nuanced way[7] than it might be with just tags because one can weigh various attributes of an artist or work of art to establish which might be the most or less important. Furthermore, such nuance allows for matching potential collectors with artworks based on their tastes and preferences.[8][9]

Artsy’s “genes” create various opportunities for discovering and learning about the artist and artworks. If users search for an artist, they can see “related” artists and if they search for an artwork, they can see “related” artworks. Genes (with definitions) also appear on their own pages and provide the backbone for Artsy’s Browse page.

The Art Genome Project provides metadata for search (and similarity) results based on the principles of information retrieval (TF/IDF) and presents results in a UX-driven search product.

Matthew Israel, an art historian, is the Director of The Art Genome Project.[10]


Precedents for The Art Genome Project

The Art Genome Project has often been compared to Pandora’s Music Genome Project, and was originally developed in consultation with Pandora’s Joe Kennedy.[11][12] Both aim to create comprehensive (though by no means exhaustive) analyses of types of art by identifying a set of criteria, which both call “genes”. These are both broadly applicable to their respective art forms, as well as useful for generating interesting connections for users. Importantly, while The Art Genome is currently one extensive list of genes for all works of art, Pandora has separate genomes (lists of genes) for each genre of music.[13]

The Music Genome Project and TAGP are examples of Big Data, which is a sector of algorithmic technology that synthesizes data to predict what users will prefer. Artsy’s genome is user facing, allowing users to navigate genes manually.[14] Netflix also makes customized recommendations to users based on the qualities of the movies that they enjoy. Amazon Art makes suggestions to users based on their browsing and purchasing history.

Museums and galleries as well as other online art image databases, such as Google Art Project and Artstor, digitize artworks for public access, but beyond providing basic metadata (artist, title, date, medium) these databases do not extensively classify works of art or create connections between them.

The Art Genome Project and Other Taxonomies for Artwork

Art historians, libraries and image archives have long used classification systems, art cataloging standards or metadata, and created taxonomies, such as The Getty Art & Architecture Thesaurus[15] to systematize the description of artworks. Classification is one of the major foundations of the discipline of art history.[16]

Most college and university art history surveys are based on such an idea of classification, to provide students with a way of grasping the history of art and a jumping-off point for more focused research. Unlike other types of classification, The Art Genome’s taxonomy was designed for the purposes of establishing similarities between artists and artworks.[17]

Genes versus Tags

In the words of Matthew Israel, Director of The Art Genome Project, “Genes are not tags — though we have many tags on the site — because tags are binary (something is either tagged “dog” or not). Genes, in contrast, can range from 0-100, thus capturing how strongly a gene applies to a specific artist or artwork. This nuanced connection between works of art is impossible with a simple tagging mechanism.”[18]

Importantly, The Art Genome Project does incorporate over 6000 tags for content (iconography) in artwork as well as certain materials and mediums, which do not need the nuance of genes.[19]


What is The Art Genome Project?

The Project has two parts.

Part 1 is a list of all of the possible characteristics/terms you might apply to art.

Think about an art object, say a painting by Andy Warhol. You might say it is a painting, that it is a work of Pop Art, that it is a silkscreen, that it features an image of Marilyn Monroe, that it is very “high contrast,” or even that it emphasizes the flatness of the image.

These characteristics or terms (e.g. Pop Art, flatness, bright colors) are what we call “genes.”

There are currently over 400 genes in what we call “The Art Genome” and they fall into the following categories. (In parentheses are examples of genes in the category.)

  • Time Period (Pre-Impressionism, Modern, Contemporary)
  • Medium (Painting, Sculpture, Installation, Video)
  • Style or Movement (Pop Art, Abstract Expressionism, Young British Artists)
  • Contemporary Tendencies (Tendencies occurring in contemporary art but that people might not yet be comfortable calling “movements,” such as Contemporary Gothic or DIY)
  • Concepts (Color Theory, Institutional Critique, Related to Film)
  • Content (Portrait, Landscape, The Studio, Cityscape)
  • Techniques (Monochrome Painting, Multiple Exposure, Sfumato)
  • Geographical Regions (Where an artist has lived and worked)
  • Appearance Genes (The look and feel of an object)
  • Labs (Genes in development; not public)

We also have hundreds of other genes. These capture individual art-historical and artist influences, such as the fact that Jackson Pollock was influenced by (among other things) Mexican Muralism or Thomas Hart Benton.

Where did all of these genes come from?

  • Hundreds of years of art-historical scholarship that we are the beneficiaries of
  • Discussions in books, periodicals and on the web surrounding contemporary art
  • Many Art.sy genome team meetings and debates
  • Consistent communication with all of our partners, i.e. the galleries, museums, foundations, collections and estates that feature their work on Art.sy.

Part 2 is applying relevant genes to each of the 3,000 artists and 15,000 artworks on Art.sy.

The list of genes applied to artists and artworks we call their “genomes.”

Like the process of coming up with genes, the application of genes to artists and artworks is a group effort, involving the genome team at Art.sy, extensive research, and consistent communication with our partners.

A few clarifications about genomes and genes:

  1. Every artist and artwork has their own genome. Why? To show how different, for example, Warhol’s oeuvre (his collected works) is in comparison to individual works and how greatly individual works can differ from each other.
  2. Genes are not tags — though we have many tags on the site — because tags are binary (something is either tagged “dog” or not). Genes, in contrast, can range from 0-100, thus capturing how strongly a gene applies to a specific artist or artwork. This nuanced connection between works of art is impossible with a simple tagging mechanism.

So this is The Art Genome Project, the source of all the terms and related searches users see on Art.sy. As always, we welcome your questions and comments.

In the coming weeks, look forward to further posts on topics like precedents for The Art Genome Project (such as art-historical taxonomies or thesauri, encyclopedias and dictionaries, image atlases, and Pandora), what appearance genes try to capture, how algorithms relate to The Art Genome Project, and how and why “Most Similar Artworks” was created.

— Matthew Israel, Director of The Art Genome Project

The Art Genome Project

The Art Genome Project is the classification system and technological framework that powers Artsy. It maps the characteristics (we call them “genes”) that connect artists, artworks, architecture, and design objects across history. There are currently over 1,000 characteristics in The Art Genome Project, including art historical movements, subject matter, and formal qualities. Read articles by The Art Genome Project, or explore our categories below.