狠狠撸

Utilization and possibilities of
art information in a data-driven society
-Characteristics of well-known Japanese oil painters
CIDOC2020
8 December,2020
Tetsuro KAMURA
Art Media Centre, Tokyo University of the Arts

Agenda
1.Introduction
2.Generate basis data and relationship indicators
3.Cluster analysis of Japanese oil painters' network
4.Analyze characteristics of well-known Japanese oil painters
5.Discussion
6.Conclusion

Background on a global level
1.The art market has been growing worldwide.
2.Open Access is implemented in the art-related field overseas.
3.Develop and using artworks or artists related authority data in
overseas museums.
4.Interdisciplinary research and information services
using relevant data of art.
5.Attempt to use diverse open data in the art market.

Issues in Japan
1. No usable artist and vocabulary data.
2. Insufficient data to purchase decisions and evaluations of
artwork.
3. A new method of valuing and evaluating artworks or artists is
required using digital data.
4. Creating new social value is required using various data in the
art field.

Objective
Reveal the importance of reusable data
infrastructure development in the art sector by
utilizing information technology and distribution
methodology.

Integration of arts and academic information Integration of arts information and business
2. Academic Research 3.Art Market
Create Innovation and Revitalizing the Art Community
Museums
Collection Info
Image Thumbnail
Provenance Info
Reference Info
Exhibition Info
Artist Info
R&D Institution
Research Data
Research method
Reference
Critics
Data from
Gallery and
Auction House
Artist, Art works
Price list
Provenance
Gallery Map
Promotion
Cataloging
Image Thumbnail
[Business]
Artworks sales and support
Information Service
Corporate patrons and funds
Appreciation of Art
Exhibit & Creation
Exhibit & Creation
Order
Statistical
Art market data
Open Data
Art & Culture data
from government
[Education & Research]
Data-driven research
Supporting Arts Education
Manage reliable art data
[Interdisciplinary approach]
Data distribution and management
Business development of study results
Support for museum field
Creation
Concept
Principle
Theme
Identity
Idea
Method
Activities
Exhibit
Communication
Research
SNS?Blog?Vlog
Analysis
Promotion
1.Artists Data
Compare artworks, price, images,
and several information.
Artists Support
Intellectual pursuit
Exhibit Planning
Art experience Consumption, Investment
Sales Services
Data
Distribution
Cycle
Infrastructure of
Art Information
Collect / Archive / Share
Acquisition of
evaluation and value
Publish
Research
Data
Use for
art business
Reviews from
public
Acquire
new knowledge
Meaning in
Culture and Society
Use for
Education
4.Public
Art Criticism
Market Data
Publish Data
The image of vitalize of the Japan’s art community
Goals
Art information has potential
use of data in the community.
We think that will be needed art-
related information pieces,
artists, and valuations based on
numerical data in the near future.
We assume building a system
that can objectively use
information owned by all
organizations and people
involved with arts.

Today’s research topics
1. Characteristics analysis using Japanese oil painters
attribute information and web services data.
2. Reveal a reality of Japanese artist valuation amount
used as one of the evaluation indicators.
3. Clarify highly rated artist elements from analysis data.

Target genre
Japanese oil painters
- Oil paintings are always in art museum globally.
- Data utilization possibilities with overseas.
If Japanese art museums publish open data

2. Generate basis data and relationship indicators
2-1. Generate raw data
2-2. Extract well-known artists
2-3. Create relationship indicators
2-4. Other components of artist value

2-1: Generate raw data
A B
Use two Japanese art yearbooks
Artist Information

Attribute Contents
Name Name of artist
Pronunciation of
kanji characters
A = Romanization
B = Katakana characters
Evaluation Price Masterpiece per the size1 (JP yen)
Association Name of art association
Alma mater University, Art school
Birthplace Prefecture
Current location Address (Residency)
Year of birth Japanese Calendar
Basic information of yearbooks
A B
Number of artists 2,476 2,958
Total(A+B) 5,434
Duplicate artists 704
Number of unique 1,772 2,254
Total
Number of unique
4,730
(704 + 1772 + 2254)
Describtive information of artist

Size
(GO)
Figure
(cm)
Paysage
(cm)
Marine
(cm)
0 17.9x13.9 17.9x11.8 17.9x10.0
1 22.1x16.6 22.1x13.9 22.1x11.8
2 24.0x19.0 24.0x16.1 24.0x13.9
3 27.3x22.0 27.3x19.0 27.3x16.1
4 33.4x24.3 33.4x21.2 33.4x19.1
5 35.0x27.3 35.0x24.3 35.0x22.1
6 40.9x21.8 40.9x27.3 40.9x24.3
8 45.5x37.9 45.5x33.3 45.5x27.3
10 53.0x45.5 53.0x40.9 53.0x33.3
12 60.6x50.0 60.6x45.5 60.6x40.9
15 65.2x53.0 65.2x50.0 65.2x45.5
20 72.7x60.6 72.7x53.0 72.7x50.0
25 80.3x65.2 80.3x60.0 80.3x53.0
30 90.9x72.7 90.9x65.2 90.9x60.6
40 100.0x80.3 100.0x72.7 100.0x65.2
50 116.7x90.9 116.7x80.3 116.7x72.7
60 130.3x97.0 130.3x89.4 130.3x80.3
80 145.5x112.1 145.5x97.0 145.5x89.4
100 162.1x130.3 162.1x112.1 162.1x97.0
120 193.9x130.3 193.9x112.1 193.9x97.0
150 227.3x181.8 227.3x162.1 227.3x145.4
200 259.1x193.9 259.1x181.8 259.1x162.1
300 290.9x218.2 290.9x197.0 290.9x181.8
500 333.3x248.5 333.3x128.2 333.3x197.0
Artist Assessment system (Known as GO system)
- We have 0 to 500 Japanese sizes standard for paintings.
- Mostly art associations or galleries evaluate a size one
price to the artist.
- Using masterpiece per the size one.
Evaluation Price = One of the assessment indicator
Example:
TANAKA‘s evaluation price = 10,000yen.
(Evaluation value = Size one price)
If created size five artwork, its price is an estimate 50,000yen.

Age Book A Book B
Blank 205 257
20 1 7
30 15 45
40 44 88
50 152 248
60 483 764
70 720 970
80 658 503
90 194 76
100 4 0
Total
2476 2958
5434(Include duplicate data)
Main layer
Evaluation
Price
Book A Book B
Blank 327 228
0 0 0
Under100,000 1345 1323
100,000 668 1320
200,000 66 45
300,000 22 15
400,000 13 7
500,000 5 7
600,000 2 3
700,000 0 2
800,000 11 0
900,000 2 2
1,000,000 9 3
2,000,000 2 3
3,000,000 2 0
4,000,000 2 0
Total 2476 2958
Compare evaluation price and age between two
books $1= 105yen, €1= 126yen (5 Dec. 2020)

2-2: Extract well-known artists
? Scraped data from Japanese art auction company’s web.
? Auction terms : 2010 to 2018.
? Auction bidding price data: 22137 records.
? Artist name text matching.
? Defined 248 artists in the marketplace as famous.
Generate data
Extracted
248 names

2-3: Create relationship indicators
- Searched a combination of each artist's names using search engine.
- Using the number of search results as a relationship indicator between artists.
Fujiwara and Yamada‘s
indicator is 62.
Indicator value of relation.
+”Fujiwara” +”Yamada”

2-4: Other components for artist valuation
Took the number of occurrences of an artist name use SPARQL API
- LODAC Museum project
130000 artworks, 8800 artists name information collected from 52 muse
- Getty ULAN LOD
- DBpedia Japanese
- Wikidata
Web Services Occurrences
LODAC Museum 109
Getty ULAN LOD 8
DBpedia Japanese 75
Wikidata 68

3. Cluster analysis of Japanese oil painters' network

A statistic of artist network
Closeness
centrality
Betweenness
Clustering
coefficient
Eigenvector
centrality
Degree
Means 0.5684 0.0032 0.7648 0.3948 57.7661
Std Error 0.0054 0.0007 0.0130 0.0190 3.2730
Median 0.5458 0.0001 0.8055 0.3367 43.5
Std Dev 0.0860 0.0112 0.2053 0.3003 51.5441
Variance 0.0074 0.0001 0.0421 0.0901 2656.8033
Min 0.3945 0 0 0.0078 1
Max 0.9047 0.1095 1 1 221
No. of Obs 248 248 248 248 248
Undirected graph, Nodes: 248, Edges: 7163
Nodes use attribute data such as evaluation prices and association.
Edges use related indicators.

Details ＼ Cluster A B C D E
Nodes 90 57 44 36 21
Edges 597 807 492 418 127
Degree(Avg.) 13 28 22 23 12
Graph Density 0.149 0.506 0.52 0.663 0.605
Evaluation(Avg.) ?250,000 ?310,000 ?200,000 ?240,000 ?600,000
Evaluation(Median) ?110,000 ?160,000 ?120,000 ?160,000 ?120,000
A B C D E
Evaluation price and
Japanese oil painters' network
? Using the Louvain algorithm
? Colour : Cluster
? Node Size: Betweenness centrality
? Label: Name – Evaluation price

Characteristics of cluster A
? The most number of nodes among the five.
? The number of edges and
the average degree are smaller than others.
? The lowest graph density value.
? Less in communication with each other.
Number of Nodes 90
Edges 597
Degree(Avg.) 13
Graph Density 0.149
Evaluation Price(Avg.) ?250,000
Evaluation Price(Median) ?110,000

Characteristics of cluster B
Number of Nodes 57
Edges 807
Degree(Avg.) 28
Graph Density 0.506
Evaluation Price(Median)? ?160,000
? The highest number of edges among the five.
? A high value of the average degree.
? The highest median evaluation price.
? A lively communication with each other.

Characteristics of cluster C
Number of Nodes 44
Edges 492
Degree(Avg.) 22
Graph Density 0.52
? Number of nodes, edges and degree are
third out of all clusters.
? A lot of artists are widely scattered.
? The lowest evaluation price of median of the five.

Characteristics of cluster D
Number of Nodes 35
Edges 417
Degree(Avg.) 23
Graph Density 0.663
? Edges and Degree are high value.
? The highest graph density of the five.
? A dense activity each other.
? The highest median evaluation price.
(Same as cluster B)

Characteristics of cluster E
? Node, Edges and Degree are lowest of all.
? Located far from the center of artist network.
? The highest average evaluation price.
? 3 and 4million yen artists in the cluster.
Number of Nodes 21
Edges 127
Degree(Avg.) 12
Graph Density 0.605

4. Analyze characteristics of well-known Japanese oil painters
Revealed elements of the evaluation price and
built a model explaining characteristics.

Categories Explanatory variables
Artworks
①Evaluation price
②Auction price
③Auction volume
Personal connection
④Alma maters
⑤Associations
Relationship indicators
⑥Closeness Centrality
⑦Betweenness
⑧Eigenvector centrality
⑨Degree
⑩Weighted degree
Web search
?Bing
?Google
Academic components
?LODAC Museum
?Getty ULAN
?DBpedia
Personal data
?Age
?Address
Build a model of evaluation price

Variable Coefficient SD t p<2.2e-16
Intercept -0.1529 0.0612 -2.496 0.013344 *
?Age × ?DBpedia 0.011 0.0033 3.46 0.000654***
?Age × ⑨Degree 7.36E-05 1.66E-05 4.443 1.43E-05***
?DBpedia × ?Address -0.0105 3.60E-03 -2.888 0.004284**
?DBpedia × ?LODAC Museum 1.408 0.2221 6.341 1.37E-09***
?DBpedia × ⑨Degree -0.0073 0.0019 -3.866 0.000147***
?LODAC Museum × ?Google search result -2.96E-05 6.54E-06 -4.526 1.00E-05***
?Google search result × ②Auction Price -1.43E-05 2.10E-06 -6.821 9.38E-11***
⑦Betweenness × ⑧Eigenvector centrality 3.28E+02 5.05E+01 6.487 6.12E-10***
⑦Betweenness × ④Alma mater -12.34 1.774 -6.955 4.34E-11***
④Alma mater × ②Auction Price 0.042 0.0038 10.927 <2.00E-16***
③Auction volume × ?Age 0.0023 0.0003 7.64 7.50E-13***
③Auction volume × ?Google 9.96E-07 4.22E-07 2.359 0.019231*
③Auction volume × ?Bing 1.59E-06 6.28E-07 2.538 0.011878*
③Auction volume × ⑨Degree 0.0023 0.0003 8.696 9.73E-16***
③Auction volume × ⑦Betweenness -2.643 0.6465 -4.089 6.17E-05***
③Auction volume × ④Association -0.0162 0.0013 -12.014 <2.00E-16***
0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.?
R2 : 0.7339, Adjust R2 : 0.7137
Create characteristics of well-known artists model using multiple regression analysis.

Relationships between the measured and predicted values of
evaluation price of artis

Excluded from the
expression
R2 Difference
②Auction price 0.5832 0.1507
③Auction volume 0.4118 0.3221
④Alma maters 0.5827 0.1512
⑤Associations 0.5518 0.1821
⑦Betweenness 0.6725 0.0614
⑧Eigenvector
centrality
0.6808 0.0531
⑨Degree 0.5563 0.1776
?Bing Search 0.7257 0.0082
?Google Search 0.6255 0.1084
?LODAC Museum 0.681 0.0529
?DBpedia 0.6615 0.0724
?Age 0.6209 0.1130
?Address 0.7233 0.0106
Most influential factors to evaluation price
Extracted variables that have a significant
impact on the coefficient of determination
and measured their importance.
Important elements
②Auction price
③Auction volume
④Alma maters
⑤Associations
⑨Degree
?Google Search
?Age
Variables affecting more than 0.1

Characteristics of high evaluation price artists.
What are the factors that makes evaluation price increase?

1. Large number of artworks widely distribute in the market.
Artworks of working artists are distributing in the market.
uThis meaning is artistic activities are likely to be active.
In short, there are ongoing production and trading.
uSomeone makes provenance, Exhibition history, and Artist information.
Biographies, reliability, and supporting information of value are on record.
uIt is advantageous to have a lot of artworks distribution.

2. Can search for artist and artwork information on the web.
A lot of information on the web.
Art auction, Exhibition, Gallery, Weblog...etc.
u Information is distributed for some purpose.
u Possibly leading to improved artist reputation and value.

3. Have multiple to connect with artists, alma mater, and associations.
Have a lots of different connections, such as various schools, art
groups, and persons.
uThat helps to get inspired by other disciplines, cultures, and
thoughts.
uIt is considered that a lots of artistic activities and experiences
are factors of increasing value and reputation.

The result of this analysis is possible to predict artists likely to
increase evaluation value in the future.
Thus, The development and utilization of various art-related data
will be able to create new social values.

Artworks,
Artist authorities
Machine Learning
Usable Data
Artists, Artworks
Recommendation
Investment proposal
Buying a piece or
investing in an artist
Personal recommendation, Trend forecasting, Trend analysis
We believe that artist introductions and trend analysis
using machine learning will be possible
when using various art-related data on the web.

The Regional trend of Japanese oil painters
Number of oil painters per 100,000 population
( ) Number of prefectures
Using National Tax Audit and Population data and 4730 Japanese oil painters’ data.
TOP10
Real number
Prefecture Person
Tokyo 933
Kanagawa 500
Saitama 350
Osaka 345
Chiba 317
Hyogo 216
Aichi 144
Ibaraki 133
Nagano 131
Kyoto 129
TOP10 Density
Per 100,000 person
Prefecture Person
Tokyo 7.05
Nara 6.47
Nagano 6.14
Kanagawa 5.51
Chiba 5.11
Kyoto 4.91
Saitama 4.85
Ibaraki 4.51
Yamanashi 3.99
Totori 3.95

Where is the most artistic
activity city ?
Algorithm: PageRank, Betweenness
Label: Pref Inflow–Outflow
Eccentricity: Green=3, Blue=4, Orange=5
TOKYO is a
center of artistic activity
In Japan
KANAGAWA is the
most of popular
residence
Top5 popular cities in
Kanagawa
1. Yokohama
2. Kamakura
3. Sagamihara
4. Kawasaki
5. Fujisawa
Using 4730 Japanese oil painters’ data.

We clarified that art-related data have the potential to use for a
variety of applications and their usefulness.
We indicated a need for concrete measures to improve the art
data infrastructure to vitalize Japan's art community.
Especially develop of following data are most important to the
art sector.
- Japanese Art thesaurus and vocabulary.
- Japanese artist authority data.
- Open / Publish artworks by usable data.
6. In conclusion

Thank you for your attention!
kamura@noc.geidai.ac.jp

狠狠撸

Utilization and possibilities of art information in a data-driven society

More Related Content

Utilization and possibilities of art information in a data-driven society