experiments in graph-based semi-supervised learning for ...partha pratim talukdar * (search labs,...

75
Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning for Class-Instance Acquisition ACL, 2010 (Uppsala, Sweden) * Work done while at the Univ. of Pennsylvania

Upload: others

Post on 16-Apr-2020

11 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Partha Pratim Talukdar* (Search Labs, MSR)

Fernando Pereira (Google)

Experiments in Graph-based Semi-Supervised Learning

for Class-Instance Acquisition

ACL, 2010 (Uppsala, Sweden) *Work done while at the Univ. of Pennsylvania

Page 2: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Class-Instance Acquisition

2

Page 3: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Class-Instance Acquisition

• Given an entity, assign human readable descriptors to it– e.g., Toyota is a car manufacturer, japanese

company, multinational company, ...

2

Page 4: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Class-Instance Acquisition

• Given an entity, assign human readable descriptors to it– e.g., Toyota is a car manufacturer, japanese

company, multinational company, ...

• Large scale, open domain

2

Page 5: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Class-Instance Acquisition

• Given an entity, assign human readable descriptors to it– e.g., Toyota is a car manufacturer, japanese

company, multinational company, ...

• Large scale, open domain

• Applications– Web search, Advertising, etc.

2

Page 6: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

3

Page 7: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

PatternTable Mining

3

Page 8: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

PatternTable Mining

3

Musician Singer

Page 9: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

PatternTable Mining

Cluster ID

3

Musician Singer

Page 10: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

PatternTable Mining

Cluster ID

ExtractionConfidence

3

Musician Singer

Page 11: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

PatternTable Mining

Cluster ID

ExtractionConfidence

3

Musician Singer

Page 12: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

SeedClasses

PatternTable Mining

Cluster ID

ExtractionConfidence

3

Musician Singer

Page 13: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition

Bob Dylan (0.95)Johnny Cash (0.87)

Billy Joel (0.82)

Set 2Set 1Billy Joel (0.75)

Johnny Cash (0.73)

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

SeedClasses

PatternTable Mining

Cluster ID

ExtractionConfidence

3

Musician Singer

Can we infer that Bob Dylan is also a Musician, as that is missing in current

extractions?

Page 14: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition [Talukdar et al., EMNLP 2008]

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

4

Page 15: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition [Talukdar et al., EMNLP 2008]

InitializationSet 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Musician 1.0

Musician 1.0

Musician 1.0

SeedLabels

4

PredictedLabels

Page 16: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition [Talukdar et al., EMNLP 2008]

Iteration 1Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Musician 1.0

Musician 0.8Musician 1.0

Musician 1.0

SeedLabels

4

PredictedLabels

Page 17: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based Class-Instance Acquisition [Talukdar et al., EMNLP 2008]

Iteration 2Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Musician 1.0

Musician 0.8Musician 1.0

Musician 1.0

SeedLabels

Musician 0.6

4

PredictedLabels

Page 18: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• Graph-based Class-Instance Acquisition– Overview & previous work

• Review & comparison of three SSL algorithms– LP-ZGL (Zhu+, 2003)

– Adsorption (Baluja+, 2008)– Modified Adsorption (MAD) (Talukdar and Crammer, 2009)

• Additional Semantic Constraints

• Summary

Outline

5

Page 19: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• Graph-based Class-Instance Acquisition– Overview & previous work

• Review & comparison of three SSL algorithms– LP-ZGL (Zhu+, 2003)

– Adsorption (Baluja+, 2008)– Modified Adsorption (MAD) (Talukdar and Crammer, 2009)

• Additional Semantic Constraints

• Summary

Outline

5

Page 20: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Comments on the Constructed Graph

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

6

Page 21: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Comments on the Constructed Graph

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Smoothness:Nodes connected by an edge should be assigned similar classes, as enforced by edge weight.

6

Page 22: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Comments on the Constructed Graph

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Smoothness:Nodes connected by an edge should be assigned similar classes, as enforced by edge weight.

Initial Clusters

6

Page 23: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Comments on the Constructed Graph

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Smoothness:Nodes connected by an edge should be assigned similar classes, as enforced by edge weight.

Initial Clusters

6

Coupling Node:Force (softly) all instance nodes connected to it to have similar class labels, exploiting the Smoothness requirement.

Page 24: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Comments on the Constructed Graph

Set 2

Set 1

Bob Dylan

Johnny Cash

Billy Joel

0.95

0.87

0.82

0.73

0.75

Musician 1.0

Musician 1.0

Smoothness:Nodes connected by an edge should be assigned similar classes, as enforced by edge weight.

Initial Clusters

Seed classes can be different from the cluster IDs

6

Coupling Node:Force (softly) all instance nodes connected to it to have similar class labels, exploiting the Smoothness requirement.

Page 25: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

LP-ZGL[Zhu et al., ICML 2003]

Page 26: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• m labels

• W : edge weight matrix

• Yul: weight of label l on node u

• Yul: seed weight of label l on node u

• S: diagonal matrix, non-zero for seed nodes

LP-ZGL[Zhu et al., ICML 2003]

Page 27: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• m labels

• W : edge weight matrix

• Yul: weight of label l on node u

• Yul: seed weight of label l on node u

• S: diagonal matrix, non-zero for seed nodes

LP-ZGL[Zhu et al., ICML 2003]

arg minY

m!

l=1

!

u,v

Wuv(Yul ! Yvl)2; s.t. SYl = SYl

Page 28: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• m labels

• W : edge weight matrix

• Yul: weight of label l on node u

• Yul: seed weight of label l on node u

• S: diagonal matrix, non-zero for seed nodes

LP-ZGL[Zhu et al., ICML 2003]

arg minY

m!

l=1

!

u,v

Wuv(Yul ! Yvl)2; s.t. SYl = SYl

Smooth

Page 29: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• m labels

• W : edge weight matrix

• Yul: weight of label l on node u

• Yul: seed weight of label l on node u

• S: diagonal matrix, non-zero for seed nodes

LP-ZGL[Zhu et al., ICML 2003]

arg minY

m!

l=1

!

u,v

Wuv(Yul ! Yvl)2; s.t. SYl = SYl

Match Seeds (hard)Smooth

Page 30: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Modified Adsorption (MAD)[Talukdar and Crammer, ECML 2009]

8

Page 31: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Modified Adsorption (MAD)[Talukdar and Crammer, ECML 2009]

8

argminY

m+1�

l=1

��SY l − SY l�2 + µ1

u,v

Muv(Y ul − Y vl)2 + µ2�Y l −Rl�2

• m labels, +1 dummy label

• M = W� +W is the symmetrized weight matrix

• Y vl: weight of label l on node v

• Y vl: seed weight for label l on node v

• S: diagonal matrix, nonzero for seed nodes

• Rvl: regularization target for label l on node v

Page 32: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Modified Adsorption (MAD)[Talukdar and Crammer, ECML 2009]

8

argminY

m+1�

l=1

��SY l − SY l�2 + µ1

u,v

Muv(Y ul − Y vl)2 + µ2�Y l −Rl�2

• m labels, +1 dummy label

• M = W� +W is the symmetrized weight matrix

• Y vl: weight of label l on node v

• Y vl: seed weight for label l on node v

• S: diagonal matrix, nonzero for seed nodes

• Rvl: regularization target for label l on node v

Match Seeds (soft)

Page 33: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Modified Adsorption (MAD)[Talukdar and Crammer, ECML 2009]

8

argminY

m+1�

l=1

��SY l − SY l�2 + µ1

u,v

Muv(Y ul − Y vl)2 + µ2�Y l −Rl�2

• m labels, +1 dummy label

• M = W� +W is the symmetrized weight matrix

• Y vl: weight of label l on node v

• Y vl: seed weight for label l on node v

• S: diagonal matrix, nonzero for seed nodes

• Rvl: regularization target for label l on node v

Match Seeds (soft) Smooth

Page 34: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Modified Adsorption (MAD)[Talukdar and Crammer, ECML 2009]

8

argminY

m+1�

l=1

��SY l − SY l�2 + µ1

u,v

Muv(Y ul − Y vl)2 + µ2�Y l −Rl�2

• m labels, +1 dummy label

• M = W� +W is the symmetrized weight matrix

• Y vl: weight of label l on node v

• Y vl: seed weight for label l on node v

• S: diagonal matrix, nonzero for seed nodes

• Rvl: regularization target for label l on node v

Match Seeds (soft) SmoothMatch Priors(Regularizer)

Page 35: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Modified Adsorption (MAD)[Talukdar and Crammer, ECML 2009]

8

argminY

m+1�

l=1

��SY l − SY l�2 + µ1

u,v

Muv(Y ul − Y vl)2 + µ2�Y l −Rl�2

• m labels, +1 dummy label

• M = W� +W is the symmetrized weight matrix

• Y vl: weight of label l on node v

• Y vl: seed weight for label l on node v

• S: diagonal matrix, nonzero for seed nodes

• Rvl: regularization target for label l on node v

MAD has extra regularization compared

to LP-ZGL

Match Seeds (soft) SmoothMatch Priors(Regularizer)

Page 36: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

MAD (contd.)

9

Inputs Y ,R : |V |× (|L|+ 1), W : |V |× |V |, S : |V |× |V | diagonalY ← YM = W +W�

Zv ← Svv + µ1�

u �=v Mvu + µ2 ∀v ∈ Vrepeat

for all v ∈ V doY v ← 1

Zv

�(SY )v + µ1Mv·Y + µ2Rv

end foruntil convergence

Page 37: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

MAD (contd.)

9

Inputs Y ,R : |V |× (|L|+ 1), W : |V |× |V |, S : |V |× |V | diagonalY ← YM = W +W�

Zv ← Svv + µ1�

u �=v Mvu + µ2 ∀v ∈ Vrepeat

for all v ∈ V doY v ← 1

Zv

�(SY )v + µ1Mv·Y + µ2Rv

end foruntil convergence

• Easily Parallelizable: Scalable

Page 38: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

MAD (contd.)

9

Inputs Y ,R : |V |× (|L|+ 1), W : |V |× |V |, S : |V |× |V | diagonalY ← YM = W +W�

Zv ← Svv + µ1�

u �=v Mvu + µ2 ∀v ∈ Vrepeat

for all v ∈ V doY v ← 1

Zv

�(SY )v + µ1Mv·Y + µ2Rv

end foruntil convergence

• Easily Parallelizable: Scalable• Importance of a node can be discounted

Page 39: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

MAD (contd.)

9

Inputs Y ,R : |V |× (|L|+ 1), W : |V |× |V |, S : |V |× |V | diagonalY ← YM = W +W�

Zv ← Svv + µ1�

u �=v Mvu + µ2 ∀v ∈ Vrepeat

for all v ∈ V doY v ← 1

Zv

�(SY )v + µ1Mv·Y + µ2Rv

end foruntil convergence

• Easily Parallelizable: Scalable• Importance of a node can be discounted• Convergence guarantee under mild conditions

Page 40: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Experiments with Public Datasets

10

Page 41: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Experiments with Public Datasets

• Previous research used proprietary datasets– difficult to reproduce and extend

10

Page 42: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Experiments with Public Datasets

• Previous research used proprietary datasets– difficult to reproduce and extend

• Public datasets are used in the paper– Freebase: relational tables from multiple sources

– TextRunner (Ritter+, 2009): Open-domain IE System

– YAGO (Suchanek+, 2007): KB curated from Wikipedia and WordNet

– Gold standard hypernyms: [Pantel+, 2009], WordNet

10

Page 43: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Experiments with Public Datasets

• Previous research used proprietary datasets– difficult to reproduce and extend

• Public datasets are used in the paper– Freebase: relational tables from multiple sources

– TextRunner (Ritter+, 2009): Open-domain IE System

– YAGO (Suchanek+, 2007): KB curated from Wikipedia and WordNet

– Gold standard hypernyms: [Pantel+, 2009], WordNet

• Available at: http://www.talukdar.net/datasets/class_inst/10

Page 44: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph Stats

11

Statistics of Graphs used in Experiments

Page 45: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Evaluation Metric: Mean Reciprocal Rank (MRR)

12

MRR =1

|test-set|�

v∈test-set

1

rankv(class(v))

Page 46: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Evaluation Metric: Mean Reciprocal Rank (MRR)

12

MRR =1

|test-set|�

v∈test-set

1

rankv(class(v))

Billy JoelLinguist 0.4Musician 0.3

...

Gold Label: MusicianMRR: 0.5 (1/2)

Page 47: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based SSL Comparisons (1)

13

Page 48: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based SSL Comparisons (1)

13

0.15

0.2

0.25

0.3

0.35

170 x 2 170 x 10

TextRunner Graph, 170 WordNet Classes

Me

an

Re

cip

roca

l R

an

k (

MR

R)

Amount of Supervision

LP-ZGL Adsorption MAD

Graph with175k nodes,529k edges.

Page 49: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Graph-based SSL Comparisons (2)

14

0.25

0.285

0.32

0.355

0.39

192 x 2 192 x 10

Freebase-2 Graph, 192 WordNet Classes

Me

an

Re

cip

roca

l R

an

k (

MR

R)

Amount of Supervision

LP-ZGL Adsorption MAD

Graph with303k nodes,2.3m edges.

Page 50: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

When is MAD most effective?

15

Page 51: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

When is MAD most effective?

15

0

0.1

0.2

0.3

0.4

0 7.5 15 22.5 30

Re

lati

ve I

ncr

eas

e i

n M

RR

by M

AD

ove

r L

P-Z

GL

Average Degree

Page 52: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

When is MAD most effective?

15

0

0.1

0.2

0.3

0.4

0 7.5 15 22.5 30

Re

lati

ve I

ncr

eas

e i

n M

RR

by M

AD

ove

r L

P-Z

GL

Average Degree

MAD is most effective in dense graphs, where there is greater need for

regularization.

Page 53: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• Graph-based Class-Instance Acquisition– Overview & previous work

• Review & comparison of three SSL algorithms– LP-ZGL (Zhu+, 2003)

– Adsorption (Baluja+, 2008)– Modified Adsorption (MAD) (Talukdar and Crammer, 2009)

• Additional Semantic Constraints

• Summary

Outline

16

Page 54: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

• Graph-based Class-Instance Acquisition– Overview & previous work

• Review & comparison of three SSL algorithms– LP-ZGL (Zhu+, 2003)

– Adsorption (Baluja+, 2008)– Modified Adsorption (MAD) (Talukdar and Crammer, 2009)

• Additional Semantic Constraints

• Summary

Outline

16

Page 55: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Can Additional Semantic Constraints Help ?

17

Page 56: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Can Additional Semantic Constraints Help ?

17

Set 2

Set 1

Isaac Newton

Johnny Cash

Bob Dylan

Page 57: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Can Additional Semantic Constraints Help ?

17

Instances with shared

attributes are likely to be

from the same class.

Set 2

Set 1

Isaac Newton

Johnny Cash

Bob Dylan

Page 58: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

has_attribute-albums

Can Additional Semantic Constraints Help ?

17

Instances with shared

attributes are likely to be

from the same class.

Set 2

Set 1

Isaac Newton

Johnny Cash

Bob Dylan

Page 59: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

has_attribute-albums

Can Additional Semantic Constraints Help ?

17

Instances with shared

attributes are likely to be

from the same class.

Set 2

Set 1

Isaac Newton

Johnny Cash

Bob Dylan

Graph-based representation makes it easy to incorporate such constraints!

Page 60: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18

Page 61: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18 0.3

0.338

0.375

0.413

0.45

170 WordNet Classes, 10 Seeds per Class, using Adsorption

Me

an

Re

cip

roca

l R

an

k (

MR

R)

TextRunner GraphYAGO GraphTextRunner + YAGO Graph

Page 62: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18 0.3

0.338

0.375

0.413

0.45

170 WordNet Classes, 10 Seeds per Class, using Adsorption

Me

an

Re

cip

roca

l R

an

k (

MR

R)

TextRunner GraphYAGO GraphTextRunner + YAGO Graph

Graph constructed from TextRunner (UWash) output, 175k nodes, 529k edges

Page 63: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18 0.3

0.338

0.375

0.413

0.45

170 WordNet Classes, 10 Seeds per Class, using Adsorption

Me

an

Re

cip

roca

l R

an

k (

MR

R)

TextRunner GraphYAGO GraphTextRunner + YAGO Graph

Graph constructed from output of YAGO Knowledge Base, 142k nodes, 777k edges

Page 64: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18 0.3

0.338

0.375

0.413

0.45

170 WordNet Classes, 10 Seeds per Class, using Adsorption

Me

an

Re

cip

roca

l R

an

k (

MR

R)

TextRunner GraphYAGO GraphTextRunner + YAGO Graph

Combined graph, with 237k nodes, 1.3m edges

Page 65: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18 0.3

0.338

0.375

0.413

0.45

170 WordNet Classes, 10 Seeds per Class, using Adsorption

Me

an

Re

cip

roca

l R

an

k (

MR

R)

TextRunner GraphYAGO GraphTextRunner + YAGO Graph

Page 66: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Better Classes with YAGO Attributes

18 0.3

0.338

0.375

0.413

0.45

170 WordNet Classes, 10 Seeds per Class, using Adsorption

Me

an

Re

cip

roca

l R

an

k (

MR

R)

TextRunner GraphYAGO GraphTextRunner + YAGO Graph

Additional semantic constraints help improve performance significantly.

Page 67: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Classes for Attributes

• Classes assigned to attribute nodes

19

has_isbn

wordnet_bookwordnet_maganize

Page 68: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Classes for Attributes

• Classes assigned to attribute nodes

19

has_isbn

wordnet_bookwordnet_maganize

Page 69: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Classes for Attributes

• Classes assigned to attribute nodes

19

has_isbn

wordnet_bookwordnet_maganize

Evidence that attributes propagate right classes.

Page 70: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Summary

20

Page 71: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Summary

• Compared three graph-based SSL algorithms for class-instance acquisition– MAD is consistently most effective, particularly in

dense graphs

20

Page 72: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Summary

• Compared three graph-based SSL algorithms for class-instance acquisition– MAD is consistently most effective, particularly in

dense graphs

• Better class-instance acquisition with additional semantic constraints– easy to add such constraints in graph-based

representation

20

Page 73: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Summary

• Compared three graph-based SSL algorithms for class-instance acquisition– MAD is consistently most effective, particularly in

dense graphs

• Better class-instance acquisition with additional semantic constraints– easy to add such constraints in graph-based

representation

• Datasets available: www.talukdar.net/datasets/class_inst/

– for reproducibility and extension

20

Page 74: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning
Page 75: Experiments in Graph-based Semi-Supervised Learning for ...Partha Pratim Talukdar * (Search Labs, MSR) Fernando Pereira (Google) Experiments in Graph-based Semi-Supervised Learning

Thank You!http://www.talukdar.net/datasets/class_inst/