Linear Coding of non-linear Hierarchies: Revitalization of an - - PowerPoint PPT Presentation

linear coding of non linear hierarchies revitalization of
SMART_READER_LITE
LIVE PREVIEW

Linear Coding of non-linear Hierarchies: Revitalization of an - - PowerPoint PPT Presentation

Problem P an .inis Solution Generalization Transfer Linear Coding of non-linear Hierarchies: Revitalization of an Ancient Classification Method Wiebke Petersen Institute of Language and Information University of Dsseldorf


slide-1
SLIDE 1

Problem P¯ an .ini’s Solution Generalization Transfer

Linear Coding of non-linear Hierarchies: Revitalization of an Ancient Classification Method

Wiebke Petersen

Institute of Language and Information University of Düsseldorf petersew@uni-duesseldorf.de

GfKl 2008

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-2
SLIDE 2

Problem P¯ an .ini’s Solution Generalization Transfer

The Problem: Sometimes we are forced to order things (nearly) linearly,

  • e. g. in . . .

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-3
SLIDE 3

Problem P¯ an .ini’s Solution Generalization Transfer

Libraries

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-4
SLIDE 4

Problem P¯ an .ini’s Solution Generalization Transfer

Warehouses

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-5
SLIDE 5

Problem P¯ an .ini’s Solution Generalization Transfer

Stores

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-6
SLIDE 6

Problem P¯ an .ini’s Solution Generalization Transfer

The problem is old!

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-7
SLIDE 7

Problem P¯ an .ini’s Solution Generalization Transfer

But more than 2 000 years ago it has been solved!

P¯ an .ini’s Grammar of Sanskrit (ca. 350 BC) Sanskrit: rich morphology, complex Sandhi-system linguistics in ancient India:

ś¯ astr¯ an¯ am . ś¯ astram ‘science of sciences’

  • ral tradition

P¯ an .ini’s grammar: system of more than 4. 000 concise rules many phonological rules

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-8
SLIDE 8

Problem P¯ an .ini’s Solution Generalization Transfer

But more than 2 000 years ago it has been solved!

P¯ an .ini’s Grammar of Sanskrit (ca. 350 BC) Sanskrit: rich morphology, complex Sandhi-system linguistics in ancient India:

ś¯ astr¯ an¯ am . ś¯ astram ‘science of sciences’

  • ral tradition

P¯ an .ini’s grammar: system of more than 4. 000 concise rules many phonological rules

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-9
SLIDE 9

Problem P¯ an .ini’s Solution Generalization Transfer

Phonological Rules

modern notation A is replaced by B if preceded by C and succeeded by D. A → B/C

D

example: final devoicing in German   + consonantal − nasal + voiced   →   + consonantal − nasal − voiced   /

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-10
SLIDE 10

Problem P¯ an .ini’s Solution Generalization Transfer

Phonological Rules

modern notation A is replaced by B if preceded by C and succeeded by D. A → B/C

D

example: final devoicing in German   + consonantal − nasal + voiced   →   + consonantal − nasal − voiced   /

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-11
SLIDE 11

Problem P¯ an .ini’s Solution Generalization Transfer

Phonological Rules

modern notation A is replaced by B if preceded by C and succeeded by D. A → B/C

D

P¯ an .ini’s linear Coding A + genitive, B + nominative, C + ablative, D + locative. example s¯ utra 6.1.77: iko yan . aci ( ) analysis: [ik]gen[yan .]nom[ac]loc modern notation: [iK] → [yN . ]/ [aC]

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-12
SLIDE 12

Problem P¯ an .ini’s Solution Generalization Transfer

Phonological Rules

modern notation A is replaced by B if preceded by C and succeeded by D. A → B/C

D

P¯ an .ini’s linear Coding A + genitive, B + nominative, C + ablative, D + locative. example s¯ utra 6.1.77: iko yan . aci ( ) analysis: [ik]gen[yan .]nom[ac]loc modern notation: [iK] → [yN . ]/ [aC]

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-13
SLIDE 13

Problem P¯ an .ini’s Solution Generalization Transfer

Śivas¯ utras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . 6. l N . 7. ñ m ˙ n n . n M 8. jh bh Ñ 9. gh d .h dh S . 10. j b g d . d Ś 11. kh ph ch t .h th c t . t V 12. k p Y 13. ś s . s R 14. h L

a.i.un . | r . .l .k |

  • e.o˙

n | ai.auc |

  • hayavarat

. | lan . | ˜ nama˙ nan . anam | jhabha˜ n | ghad . hadhas . | jabagad . ada´ s | khaphachat .hathacat .atav | kapay | ´ sas .asar | hal |

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-14
SLIDE 14

Problem P¯ an .ini’s Solution Generalization Transfer

Śivas¯ utras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . 6. l N . 7. ñ m ˙ n n . n M 8. jh bh Ñ 9. gh d .h dh S . 10. j b g d . d Ś 11. kh ph ch t .h th c t . t V 12. k p Y 13. ś s . s R 14. h L

a.i.un . | r . .l .k |

  • e.o˙

n | ai.auc |

  • hayavarat

. | lan . | ˜ nama˙ nan . anam | jhabha˜ n | ghad . hadhas . | jabagad . ada´ s | khaphachat .hathacat .atav | kapay | ´ sas .asar | hal |

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-15
SLIDE 15

Problem P¯ an .ini’s Solution Generalization Transfer

Śivas¯ utras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . 6. l N . 7. ñ m ˙ n n . n M 8. jh bh Ñ 9. gh d .h dh S . 10. j b g d . d Ś 11. kh ph ch t .h th c t . t V 12. k p Y 13. ś s . s R 14. h L marker

a.i.un . | r . .l .k |

  • e.o˙

n | ai.auc |

  • hayavarat

. | lan . | ˜ nama˙ nan . anam | jhabha˜ n | ghad . hadhas . | jabagad . ada´ s | khaphachat .hathacat .atav | kapay | ´ sas .asar | hal |

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-16
SLIDE 16

Problem P¯ an .ini’s Solution Generalization Transfer

Śivas¯ utras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . 6. l N . 7. ñ m ˙ n n . n M 8. jh bh Ñ 9. gh d .h dh S . 10. j b g d . d Ś 11. kh ph ch t .h th c t . t V 12. k p Y 13. ś s . s R 14. h L marker

a.i.un . | r . .l .k |

  • e.o˙

n | ai.auc |

  • hayavarat

. | lan . | ˜ nama˙ nan . anam | jhabha˜ n | ghad . hadhas . | jabagad . ada´ s | khaphachat .hathacat .atav | kapay | ´ sas .asar | hal |

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-17
SLIDE 17

Problem P¯ an .ini’s Solution Generalization Transfer

praty¯ ah¯ aras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T .

a praty¯ ah¯ ara is a pair of a sound and a marker it denotes the continuous sequence of sounds in the interval between the sound and the marker

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-18
SLIDE 18

Problem P¯ an .ini’s Solution Generalization Transfer

praty¯ ah¯ aras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . iK

a praty¯ ah¯ ara is a pair of a sound and a marker it denotes the continuous sequence of sounds in the interval between the sound and the marker

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-19
SLIDE 19

Problem P¯ an .ini’s Solution Generalization Transfer

praty¯ ah¯ aras

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . iK= i, u, r ., l .

a praty¯ ah¯ ara is a pair of a sound and a marker it denotes the continuous sequence of sounds in the interval between the sound and the marker

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-20
SLIDE 20

Problem P¯ an .ini’s Solution Generalization Transfer

Analysis of iko yan .aci: [iK] → [yN . ]/ [aC]

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . 6. l N .

[iK] → [yN . ]/ [aC] i, u, r ., l . → y, v, r, l/ a, i, u, r ., l ., e, o, ai, au

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-21
SLIDE 21

Problem P¯ an .ini’s Solution Generalization Transfer

Analysis of iko yan .aci: [iK] → [yN . ]/ [aC]

1. a i u N . 2. r . l . K 3. e

  • ˙

N 4. ai au C 5. h y v r T . 6. l N .

[iK] → [yN . ]/ [aC] i, u, r ., l . → y, v, r, l/ a, i, u, r ., l ., e, o, ai, au

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-22
SLIDE 22

Problem P¯ an .ini’s Solution Generalization Transfer

Generalized task

task Given a set of classes, order the elements of the classes linearly such that each class forms an interval. If unavoidable, duplicate some elements, but minimize the number of duplications.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-23
SLIDE 23

Problem P¯ an .ini’s Solution Generalization Transfer

Generalized task

task Given a set of classes, order the elements of the classes linearly such that each class forms an interval. If unavoidable, duplicate some elements, but minimize the number of duplications.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-24
SLIDE 24

Problem P¯ an .ini’s Solution Generalization Transfer

Terminology: S-sortability

set of classes (A, Φ): A = {a, b, c, d, e, f , g, h, i}} Φ = {{d, e}, {a, b}, {b, c, d, f , g, h, i}, {f , i}, {c, d, e, f , g, h, i}, {g, h}} S-order (A <) of (A, Φ): a b c g h f i d e

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-25
SLIDE 25

Problem P¯ an .ini’s Solution Generalization Transfer

Terminology: S-sortability

set of classes (A, Φ): A = {a, b, c, d, e, f , g, h, i}} Φ = {{d, e}, {a, b}, {b, c, d, f , g, h, i}, {f , i}, {c, d, e, f , g, h, i}, {g, h}} S-order (A <) of (A, Φ): a b c g h f i d e

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-26
SLIDE 26

Problem P¯ an .ini’s Solution Generalization Transfer

Terminology: S-sortability

set of classes (A, Φ): A = {a, b, c, d, e, f , g, h, i}} Φ = {{d, e}, {a, b}, {b, c, d, f , g, h, i}, {f , i}, {c, d, e, f , g, h, i}, {g, h}} S-order (A <) of (A, Φ): a b c g h f i d e

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-27
SLIDE 27

Problem P¯ an .ini’s Solution Generalization Transfer

Terminology: S-sortability

set of classes (A, Φ): A = {a, b, c, d, e, f , g, h, i}} Φ = {{d, e}, {a, b}, {b, c, d, f , g, h, i}, {f , i}, {c, d, e, f , g, h, i}, {g, h}} S-order (A <) of (A, Φ): a b c g h f i d e ⇒ (A, Φ) is S-sortable without duplications

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-28
SLIDE 28

Problem P¯ an .ini’s Solution Generalization Transfer

Terminology: S-sortability

set of classes (A, Φ): A = {a, b, c, d, e, f , g, h, i}} Φ = {{d, e}, {a, b}, {b, c, d, f , g, h, i}, {f , i}, {c, d, e, f , g, h, i}, {g, h}}

e d c i f h g b a {d, e} {d} {c, d, f , g, h, i} {f , i} {g, h} {b} {a, b} { } {a, b, c, d, e, f , g, h, i} {c, d, e, f , g, h, i} {b, c, d, f , g, h, i}

a b c d e f g h i {d, e} ×× {b, c, d, f , g, h, i} ×××××××× {a, b} ×× {f , i} × × {c, d, e, f , g, h, i} ××××××× {g, h} ××

concept lattice of (A, Φ) formal context of (A, Φ)

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-29
SLIDE 29

Problem P¯ an .ini’s Solution Generalization Transfer

Terminology: S-sortability

set of classes (A, Φ): A = {a, b, c, d, e, f , g, h, i}} Φ = {{d, e}, {a, b}, {b, c, d, f , g, h, i}, {f , i}, {c, d, e, f , g, h, i}, {g, h}}

e d c i f h g b a

a b c d e f g h i {d, e} ×× {b, c, d, f , g, h, i} ×××××××× {a, b} ×× {f , i} × × {c, d, e, f , g, h, i} ××××××× {g, h} ××

concept lattice of (A, Φ) formal context of (A, Φ)

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-30
SLIDE 30

Problem P¯ an .ini’s Solution Generalization Transfer

Determining S-orders

e d c i f h g b a

Theorem A set of classes (A, Φ) is S-sortable without duplications iff its concept lattice is a planar graph and for any a ∈ A there is a node labeled a in the S-graph.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-31
SLIDE 31

Problem P¯ an .ini’s Solution Generalization Transfer

Determining S-orders

e d c i f h g b a

Theorem A set of classes (A, Φ) is S-sortable without duplications iff its concept lattice is a planar graph and for any a ∈ A there is a node labeled a in the S-graph.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-32
SLIDE 32

Problem P¯ an .ini’s Solution Generalization Transfer

Determining S-orders

e d c i f h g b a

Theorem A set of classes (A, Φ) is S-sortable without duplications iff its concept lattice is a planar graph and for any a ∈ A there is a node labeled a in the S-graph.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-33
SLIDE 33

Problem P¯ an .ini’s Solution Generalization Transfer

Determining S-orders

e d c i f h g b a

e d c f i g h b a

Theorem A set of classes (A, Φ) is S-sortable without duplications iff its concept lattice is a planar graph and for any a ∈ A there is a node labeled a in the S-graph.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-34
SLIDE 34

Problem P¯ an .ini’s Solution Generalization Transfer

Enlarged praty¯ ah¯ ara-concept-lattice

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-35
SLIDE 35

Problem P¯ an .ini’s Solution Generalization Transfer

Enlarged praty¯ ah¯ ara-concept-lattice

By duplicating elements all sets of classes become S-sortable!

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-36
SLIDE 36

Problem P¯ an .ini’s Solution Generalization Transfer

Main theorem of S-sortability

A set of classes (A, Φ) is S-sortable without duplications if one of the following equivalent statements is true:

1

The concept lattice of (A, Φ) is a Hasse-planar graph and for any a ∈ A there is a node labeled a in the S-graph.

2

The concept lattice of the enlarged set of classes (A, ˜ Φ) is Hasse-planar. (˜ Φ = Φ ∪ {{a}

  • a ∈ A})

3

The Ferrers-graph of the enlarged (A, ˜ Φ)-context is bipartite.

Example: S-sortable

e d c i f h g b a

Example: not S-sortable

d c b e f a

{{d, e}, {a, b}, {b, c, d}, {b, c, d, f }}

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-37
SLIDE 37

Problem P¯ an .ini’s Solution Generalization Transfer

Main theorem of S-sortability

A set of classes (A, Φ) is S-sortable without duplications if one of the following equivalent statements is true:

1

The concept lattice of (A, Φ) is a Hasse-planar graph and for any a ∈ A there is a node labeled a in the S-graph.

2

The concept lattice of the enlarged set of classes (A, ˜ Φ) is Hasse-planar. (˜ Φ = Φ ∪ {{a}

  • a ∈ A})

3

The Ferrers-graph of the enlarged (A, ˜ Φ)-context is bipartite. Example: S-sortable

e d c i f h g b a

Example: not S-sortable

d c b e f a

{{d, e}, {a, b}, {b, c, d}, {b, c, d, f }}

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-38
SLIDE 38

Problem P¯ an .ini’s Solution Generalization Transfer

Main theorem of S-sortability

A set of classes (A, Φ) is S-sortable without duplications if one of the following equivalent statements is true:

1

The concept lattice of (A, Φ) is a Hasse-planar graph and for any a ∈ A there is a node labeled a in the S-graph.

2

The concept lattice of the enlarged set of classes (A, ˜ Φ) is Hasse-planar. (˜ Φ = Φ ∪ {{a}

  • a ∈ A})

3

The Ferrers-graph of the enlarged (A, ˜ Φ)-context is bipartite.

Example: S-sortable

e d c i f h g b a

e d c i f h g b a

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-39
SLIDE 39

Problem P¯ an .ini’s Solution Generalization Transfer

Main theorem of S-sortability

A set of classes (A, Φ) is S-sortable without duplications if one of the following equivalent statements is true:

1

The concept lattice of (A, Φ) is a Hasse-planar graph and for any a ∈ A there is a node labeled a in the S-graph.

2

The concept lattice of the enlarged set of classes (A, ˜ Φ) is Hasse-planar. (˜ Φ = Φ ∪ {{a}

  • a ∈ A})

3

The Ferrers-graph of the enlarged (A, ˜ Φ)-context is bipartite. Example: not S-sortable

d c b e f a d c b e f a

{{d, e}, {a, b}, {b, c, d}, {b, c, d, f }}

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-40
SLIDE 40

Problem P¯ an .ini’s Solution Generalization Transfer

Main theorem of S-sortability

A set of classes (A, Φ) is S-sortable without duplications if one of the following equivalent statements is true:

1

The concept lattice of (A, Φ) is a Hasse-planar graph and for any a ∈ A there is a node labeled a in the S-graph.

2

The concept lattice of the enlarged set of classes (A, ˜ Φ) is Hasse-planar. (˜ Φ = Φ ∪ {{a}

  • a ∈ A})

3

The Ferrers-graph of the enlarged (A, ˜ Φ)-context is bipartite. Advantages: The Ferrers-graph is constructed

  • n the formal context.

Its bipartity can be checked algorithmically.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-41
SLIDE 41

Problem P¯ an .ini’s Solution Generalization Transfer

Between trees and general hierarchies

tree S-sortable general hierarchy

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-42
SLIDE 42

Problem P¯ an .ini’s Solution Generalization Transfer

Transfer

For physical objects ‚duplicating‘ means ‚adding copies‘ Adding copies is annoying but often not impossible Ordering objects in an S-order may

improve user-friendliness save time save space simplify visual representations of classifications

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-43
SLIDE 43

Problem P¯ an .ini’s Solution Generalization Transfer

Transfer

For physical objects ‚duplicating‘ means ‚adding copies‘ Adding copies is annoying but often not impossible Ordering objects in an S-order may

improve user-friendliness save time save space simplify visual representations of classifications

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-44
SLIDE 44

Problem P¯ an .ini’s Solution Generalization Transfer

Transfer

For physical objects ‚duplicating‘ means ‚adding copies‘ Adding copies is annoying but often not impossible Ordering objects in an S-order may

improve user-friendliness save time save space simplify visual representations of classifications

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-45
SLIDE 45

Problem P¯ an .ini’s Solution Generalization Transfer

Outlook

Objects in libraries, ware-houses, and stores are only nearly linearly arranged: ⇒ Second (and third) dimension can be used in order to avoid duplications

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-46
SLIDE 46

Problem P¯ an .ini’s Solution Generalization Transfer

Literature

Böhtlingk, O. (1887), P¯ an . inis Grammatik. Leipzig, Nachdruck Hildesheim 1964. Ganter, B. & R. Wille (1996), Formale Begriffsananlyse - Mathematische

  • Grundlagen. Berlin: Springer.

Kiparsky, P. (1991), Economy and the construction of the Śivas¯

  • utras. In:
  • M. M. Deshpande & S. Bhate (Ed.), P¯

an . inian Studies, Michigan: Ann Arbor, (Wieder abgedruckt auf: http://www.stanford.edu/~kiparsky/Papers/siva-t.pdf, 15 Seiten). Kiparsky, P. (2002), On the architecture of P¯ an .ini’s grammar, three lectures delivered at the Hyderabad Conference on the Architecture of Grammar, Jan. 2002, and at UCLA, March 2002 (http://www.stanford.edu/~kiparsky/Papers/hyderabad.pdf). Petersen, W. (2008), Zur Minimalität von P¯ an .inis Śivas¯ utras – Eine Untersuchung mit Mitteln der Formalen Begriffsanalyse. PhD thesis, university of Düsseldorf.

Linear Coding of non-linear Hierarchies Wiebke Petersen

slide-47
SLIDE 47

Problem P¯ an .ini’s Solution Generalization Transfer

Origin of Pictures

libraries (left): http://www.meduniwien.ac.at/medizinischepsychologie/bibliothek.htm libraries (middle): http://www.math-nat.de/aktuelles/allgemein.htm libraries (right): http://www.geschichte.mpg.de/deutsch/bibliothek.html warehouses: http://www.metrogroup.de/servlet/PB/menu/1114920_l1/index.html stores: http://www.einkaufsparadies-schmidt.de/01bilder01/

Linear Coding of non-linear Hierarchies Wiebke Petersen