Computerized classification of large ceramic assemblages

A quantitative basis for relative chronology

Avshalom Karasik

The Computerized Archaeological Laboratory, Hebrew University of Jerusalem

Who are we?

Uzy Smilansky

Leore Grosman

Talia Goldman Avshalom Karasik

Ilan Sharon

Ortal Haroch

"Pottery is, however, the greatest resource of the archaeologist. For variety of form and texture, for decoration, for rapid change, for its quick fall into oblivion, and for its incomparable abundance, it is in every respect the most important material for study, and it constitutes the essential alphabet of archaeology in every land.…to tie together sequences found at related sites in a region to form a master chronological sequence. This would enable any absolute dates determined from one site (for example through inscriptions, documentary evidence, and so on) to be transferred to other sites in the master sequence”

F. Petrie 1904

"Pottery is, however, the greatest resource of the archaeologist. For variety of form and texture, for decoration, for rapid change, for its quick fall into oblivion, and for its incomparable abundance, it is in every respect the most important material for study, and it constitutes the essential alphabet of archaeology in every land.…to tie together sequences found at related sites in a region to form a master chronological sequence. This would enable any absolute dates determined from one site (for example through inscriptions, documentary evidence, and so on) to be transferred to other sites in the master sequence”

F. Petrie 1904

A Test Case – Torpedo Storage Jars

The assemblage: Torpedo storage jars from Hazor (53) and Tyre (24), that have been already discussed in the literature in regards to their similarity and its historical consequences.

• Geva S. BASOR 248 Pp 69-72. 1982. Bikai P. BASOR 258 Pp 71-72. 1985. Gilboa A. In QEDEM Reports 2, page 11. 1995.

.  א , - לאמר אלי יהוה דבר -בויהי , - על שא אדם בן .  צרואתה , גקינה שבת הי לצור ואמרת , , ,  ; - , , כלילת- אני אמרת את צור יהוה אדני אמר ה כ רבים ים אי אל העמים רכלת ים מבואת על

, יזיפי.  ...  רכליך המה ישראל וארץ , יהודה , , ; נתנו וצרי ושמן ודבש ופנג מנית בחטי מערבך. 

The word of the Lord came to me: 2Now you, mortal, raise a lamentation over Tyre, 3and say to Tyre, which sits at the entrance to the sea, merchant of the peoples on many coastlands, Thus says the Lord God: O Tyre, you have said, ‘I am perfect in beauty.’ … 17Judah and the land of Israel traded with you; they exchanged for your merchandise wheat from Minnith, millet,* honey, oil, and balm.

Ezekiel 27

A Test Case – Torpedo Storage Jars

A correlation matrix

0.4

0.5

0.6

0.7

0.8

0.9

1

10 20 30 40 50 60 70

10

20

30

40

50

60

70

Hazor

Tyre

Jar Index

Jar Index

A Test Case – Torpedo Storage Jars

214143 91018425348202529542669386162707727724965687445735556576359586466606771 1 4 51475 2 7223932 3362335407644 837503446 6473051131719312833111624521215-5

0

5

10

15

20

25

30

35

40

45

A Test Case – Torpedo Storage Jars

• Tyre• Hazor

• There is a significant morphological difference between the jars from Tyre and those from Hazor.

• Any theory that describes the commercial connections between the two sites based on the similarity of the jars must explain this independent dichotomy.

• For instance: it is possible that the differences are due to the fact that the two assemblages were drawn by two different draftspersons.

• The archaeological publications must progress to automatic digital documentation of ceramic.

A Test Case – Torpedo Storage Jars: Conclusions

High precision data acquisition using 3D scanners.

We have developed an innovative algorithm which automatically finds the symmetry axis of pottery fragments and extracts their profiles.

Final products – automatically produced print quality drawings

Final products – automatically produced print quality drawings with colored views

0 10 20 cm

0 70 1400

40

80

X

Y

-100 -50 0 50 100110

115

120

125

130

135

140

Arc-length

Rad

ius

-100 -50 0 50 1000

1

2

3

4

5

6

Arc-length

Tang

ent

-100 -50 0 50 100-0.2

-0.1

0

0.1

0.2

0.3

Arc-length

Cur

vatu

re

=

Automatic classification and typology of ceramic fragments

0 70 1400

40

80

X

Y

-100 -50 0 50 100100

110

120

130

140

Arc-length

Rad

ius

-100 -50 0 50 1000

1

2

3

4

5

6

Arc-length

Tang

ent

-100 -50 0 50 100-0.2

-0.1

0

0.1

0.2

0.3

Arc-length

Cur

vatu

re

Automatic classification and typology

0 70 1400

40

80

X

Y

-100 -50 0 50 100100

110

120

130

140

Arc-length

Rad

ius

-100 -50 0 50 1000

1

2

3

4

5

6

Arc-length

Tang

ent

-100 -50 0 50 100-0.2

-0.1

0

0.1

0.2

0.3

Arc-length

Cur

vatu

re

Automatic classification and typology

0 70 1400

40

80

X

Y

-100 -50 0 50 100110

115

120

125

130

135

140

Arc-length

Rad

ius

-100 -50 0 50 1000

1

2

3

4

5

6

Arc-length

Tang

ent

-100 -50 0 50 100-0.2

-0.1

0

0.1

0.2

0.3

Arc-length

Cur

vatu

re

Automatic classification and typology

8 10 12 14 16 18

1

2

3

4

5

6

Bowl

12 14 16 18 20 2218

19

20

21

22

23

24Krater

8 10 12 14 16 182

3

4

5

6

7

8Cooking pot

0 2 4 6 8 106

7

8

9

10

11

12Jug

0 2 4 6 8 10 1223

24

25

26

27

28

29Jar

-10 -8 -6 -4 -2 0 2 4 6 8 102

4

6

8

10

12

14

16

18

20

Arc-length

Rad

ius

BowlKraterCooking potJugJar

-10 -8 -6 -4 -2 0 2 4 6 8 100

1

2

3

4

5

6

Arc-length

Tang

ent

BowlKraterCooking potJugJar

-10 -8 -6 -4 -2 0 2 4 6 8 10-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Arc-length

Cur

vatu

re

BowlKraterCooking potJugJar

Automatic classification and typology

Fragment Index

Frag

men

t Ind

ex

100 200 300

100

200

300

0

0.5

1

1.5

2Distance Matrix

Automatic classification and typology

0

500

1000

1500

2000

2500

3000

3500

Jars and Jugs

Open bowls

variacraters

Cooking pots and Craters

Primary clustering: Example: 755 Iron-age fragments from area G in Tel-Dor (Israel)

0

500

1000

1500

2000

2500

3000

3500

0 10 20 cm

0 10 20 cm

0 10 20 cm

122123110

120

100 300

200

213110

220

212

221 222

214 211 311313 314 312121

321 322315

330

230

230 210330310 320

Detailed clustering: Example: 380 Iron-age fragments from area G in Tel-Dor (Israel), that were already analyzed by Ayelet Gilboa (Haifa University).

110 121 122 123 211 212 213 214 221 222 230 311 312 313 314 315 321 322 3300

10

20

30

40

50

60

70

Type-Code

Num

ber

of v

esse

ls

BowlsKratersCooking potsJugsJars

Detailed clustering: Example: 380 Iron-age fragments from area G in Tel-Dor (Israel), that were already analyzed by Ayelet Gilboa (Haifa University).

The 300 subtypes

0 10 20 cm

1 2 3 4 5 6 7 8 9 10

11 12 13 14 15 16 17 18 19 20 21 22

23 24 25 26 27 28 29

30 31 32 33 34 35 36 37 38 39

Type-Code 311 Type-Code 313Type-Code 312

Type-Code 314 Type-Code 315

Type-Code 321

Type-Code 330Type-Code 322

Detailed clustering: Example: 380 Iron-age fragments from area G in Tel-Dor (Israel), that were already analyzed by Ayelet Gilboa (Haifa University).

-50 -40 -30 -20 -10 0 10

-0.1

-0.05

0

0.05

0.1

Arc-length (mm)

k(s)

Type-code 211Type-code 212Type-code 213Type-code 214

type-code 211type-code 212

type-code 213 type-code 2140 10 20 cm

Discriminant Analysis of the results

Original type-codes

DA

labe

ling

100 200 300

100

200

300

0 10 20 cm

0 10 20 cm

110 120

120

110

310 320 330

330

320

310

Original type-codes

DA

labe

ling

210 220 230

230

220

210

Kefar Hananya : Late Hellenistic - early Roman pottery production center. Can one distinguish systematic style variations between different workshops in the same village?

High resolution typology

Bowls type 1E

Cooking pots type 4C

In collaboration with Prof. David Adan-Bayewich (Bar Ilan University)

E1 bowls

High resolution typology

0

20

40

60

80

100

Clu

ster

hei

ght

Main excavation areaSquare O4

1 2

1b 1a

3

Discriminant Analysis

0 0.5 10

10

20

30

40

50

60

70

Provenance Index

%

Square O4Main excavation area

Chemical analysis10 element abundances measured by neutron

activation and high-precision X-ray fluorescence analyses

Shape analysis

Summary and Conclusions

This presentation describes a novel method for computerized ceramic typology and classification which is based on a numeric distance function between three mathematical representations of the profile.

Automatic classification can follow the traditional archaeological concepts of typological classification and even improve its resolution.

The classification procedure is fast, objective, reproducible and has no bias from subjective judgments.

Computerized classification of large ceramic assemblages can and should serve as a quantitative basis for relative chronology.

Thank you for your attention!

122123110

120

100 300

200

213110

220

212

221 222

214 211 311313 314 312121

321 322315

330

230

230 210330310 320

110 121 122 123 211 212 213 214 221 222 230 311 312 313 314 315 321 322 3300

10

20

30

40

50

60

70

Type-Code

Am

ount

BowlsKratersCooking potsJugsJars

0 70 1400

40

80

X

Y

-100 -50 0 50 100100

110

120

130

140

Arc-length

Rad

ius

-100 -50 0 50 1000

1

2

3

4

5

6

Arc-length

Tang

ent

-100 -50 0 50 100-0.2

-0.1

0

0.1

0.2

0.3

Arc-lengthC

urva

ture