1 minimally supervised morphological analysis by multimodal alignment david yarowsky and richard...

29
1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

Upload: aliza-hogarth

Post on 01-Apr-2015

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

1

Minimally Supervised Morphological Analysis by

Multimodal Alignment

David Yarowsky and

Richard Wicentowski

Page 2: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

2

Introduction

The Algorithm capable of inducing inflectional morphological analyses of regular and highly irregular forms.

The Algorithm combines four original alignment models based on:• Relative corpus frequency.

• Contextual Similarity.

• Weighted string similarity.

• Incrementally retrained inflectional transduction probabilities.

Page 3: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

3

Lecture’s Subjects

Task definition. Required and Optional resources. The Algorithm. Empirical Evaluation.

Page 4: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

4

Task Definition

Consider this task as three steps:• Estimate a probabilistic alignment between

inflected forms and root forms.

• Train a supervised morphological analysis learner on a weighted subset of these aligned pairs.

• Use the result from step 2 to iteratively refine the alignment in step 1.

Page 5: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

5

Example (POS)

Definitions:

Page 6: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

6

Task Definition cont.

The target output of step 1:

Page 7: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

7

Required and Optional resources

For the given language we need:• A table of the inflectional Part of Speech

(POS).

• A list of the canonical suffixes.

A large text corpus.

Page 8: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

8

Required and Optional resources cont.

A list of the candidate noun, verb and adjective roots (from dictionary), and any rough mechanism for identifying the candidates POS of the remaining vocabulary. (not based on morphological analysis).

A list of the consonants and vowels.

Page 9: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

9

Required and Optional resources cont.

A list of common function words.

A distance/similarity tables generated on previously studied languages.

Not essential

If available

Page 10: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

10

The Algorithm

Combines four original alignment models:• Alignment by Frequency Similarity.

• Alignment by Context Similarity.

• Alignment by Weighted Levenshtein Distance.

• Alignment by Morphological Transformation Probabilities.

Page 11: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

11

Lemma Alignment by Frequency Similarity

The motivating dilemma:

sing singedVBD

?

sing sangVBD

?

take takedVBD

?

Page 12: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

12

Lemma Alignment by Frequency Similarity cont.

This Table is based on relative corpus frequency:

Page 13: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

13

Lemma Alignment by Frequency Similarity cont.

Page 14: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

14

Lemma Alignment by Frequency Similarity cont.

A problem: the true alignments between inflections are unknown in advance.

A simplifying assumption: the frequency ratios between inflections and roots is not significantly different between regular and irregular morphological processes.

Page 15: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

15

Lemma Alignment by Frequency Similarity cont.

Similarity between regular and irregular forms:

Page 16: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

16

Lemma Alignment by Frequency Similarity cont.

The expected frequency should also be estimable from the frequency of any of the other inflectional variants.

VBD/VBG and VBD/VBZ could also be used as estimators.

Page 17: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

17

Lemma Alignment by Frequency Similarity cont.

Page 18: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

18

Lemma Alignment by Context Similarity

Based on contextual similarity of the candidate form.

Computing similarity between vectors of weighted and filtered context features.

Clustering inflectional variants of verbs (e.g. sipped, sipping, and sip).

Page 19: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

19

Lemma Alignment by Context Similarity cont.

Example:

CWsubj(AUX|NEG)*VkeywordDET?CW*CWobj

eating the appleShlomo is

Page 20: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

20

Lemma Alignment by Weighted Levenshtein Distance

Consider overall stem edit distance. A cost matrix with initial distance costs:

initially set to (0.5,0.6,1.0,0.98)

VCCCVVVV 4321 ,,,

Page 21: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

21

Lemma Alignment by Morphological Transformation Probabilities

The goal is to generalize a mapping function via a generative probabilistic model.

Page 22: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

22

Lemma Alignment by Morphological Transformation Probabilities

Result table:

Page 23: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

23

Lemma Alignment by Morphological Transformation Probabilities cont.

<root>+<stem change>+<suffix><inflection>

P(inflection | root,suffix,POS)=P(stemchange | root,suffix,POS)

unique

Page 24: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

24

Lemma Alignment by Morphological Transformation Probabilities cont.

),,|(),,|(

,,

POSPPOSP

inflectionsuffixroot

Example:

Page 25: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

25

Lemma Alignment by Morphological Transformation Probabilities cont.

Example:

P(solidified | solidify, +ed, VBD)

= P(yi | solidify, +ed, VBD)

≈ 1P(yi | ify, +ed)

+ (1-1)( 2P(yi | fy, +ed)

+ (1-2)( 3P(yi | y, +ed)

+ (1-3)( 4P(yi | +ed)

+ (1-4) P(yi)

POS can be deleted

Page 26: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

26

Lemma Alignment by Model Combination and the Pigeonhole Principle

No single model is sufficiently effective on its own.

The Frequency, Levenshtein and Context Similarity models retain equal relative weight.

The Morphological Transformation Similarity model increases in relative weight.

Page 27: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

27

Lemma Alignment by Model Combination and the Pigeonhole Principle

Example:

Page 28: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

28

Lemma Alignment by Model Combination and the Pigeonhole Principle cont.

The final alignment is based on the pigeonhole principle.

For a given POS a root shouldn't have more than one inflection nor should multiple inflections in the same POS share the same root.

Page 29: 1 Minimally Supervised Morphological Analysis by Multimodal Alignment David Yarowsky and Richard Wicentowski

29

Empirical Evaluation

Performance: