Scalable Learning and Inference in Large Knowledge Bases Yang Chen - - PowerPoint PPT Presentation

▶
scalable learning and inference in large knowledge bases
SMART_READER_LITE
LIVE PREVIEW

Scalable Learning and Inference in Large Knowledge Bases Yang Chen - - PowerPoint PPT Presentation

Aug 16, 2023 50 likes •1.1k views

Scalable Learning and Inference in Large Knowledge Bases Yang Chen Data Science Research Lab University of Florida @ D ata S cience R esearch Agenda Ontological Knowledge Introduction Conclusion Pathfinding Expansion Agenda Ontological

slide-1
SLIDE 1

Scalable Learning and Inference in Large Knowledge Bases

Yang Chen

Data Science Research Lab University of Florida

Data Science Research

@

slide-2
SLIDE 2

Agenda

Introduction Ontological Pathfinding Knowledge Expansion Conclusion

slide-3
SLIDE 3

Agenda

Introduction Ontological Pathfinding Knowledge Expansion Conclusion SIGMOD’16

slide-4
SLIDE 4

Agenda

Introduction Ontological Pathfinding Knowledge Expansion Conclusion VLDBJ’16 SIGMOD’14 VLDB’16

slide-5
SLIDE 5

Introduction

slide-6
SLIDE 6

“Science is built up

  • f facts, as a house is

with stones.”

— Jules Henri Poincar´e La Science et l’Hypoth`ese, 1901

slide-7
SLIDE 7

Knowledge Bases

BornIn LiveIn LiveIn BornIn LocatedIn

slide-8
SLIDE 8
slide-9
SLIDE 9

Knowledge Base Construction

slide-10
SLIDE 10

Knowledge Base Construction

ProbKB

slide-11
SLIDE 11

Incomplete Knowledge in Freebase

Known Place of Birth 30%

Unknown Place of Birth 70%

slide-12
SLIDE 12

Incomplete Knowledge in Freebase

Known ethnicity 1%

Unknown ethnicity 99%

slide-13
SLIDE 13

Question

Can we expand incomplete, uncertain knowledge bases by first-order inference?

slide-14
SLIDE 14

First-Order Inference

BornIn LiveIn LiveIn BornIn LocatedIn

slide-15
SLIDE 15

First-Order Inference

BornIn LiveIn LiveIn BornIn LocatedIn BornIn(x, z), LocatedIn(z, y) BornIn(x, y)

slide-16
SLIDE 16

First-Order Inference

BornIn LiveIn LiveIn BornIn LocatedIn wasBornIn LocatedIn

slide-17
SLIDE 17

Related Work

Learning Sherlock (OpenIE) AMIE+ (YAGO) Inference NELL DeepDive (Tuffy)

slide-18
SLIDE 18

The Open World Assumption

BornIn LiveIn LiveIn BornIn LocatedIn studiedIn

?

slide-19
SLIDE 19

AMIE+

Runtime/min 125 250 375 500 Input size 3000000 6000000 9000000 12000000 YAGO YAGO2s DBpedia

Input Size Runtime/min

slide-20
SLIDE 20

AMIE+

Runtime/min 125 250 375 500 Input size 100000000 200000000 300000000 400000000 YAGO YAGO2s DBpedia Freebase

Input Size Runtime/min

Better Scalability?

slide-21
SLIDE 21

Related Work

Learning Sherlock (OpenIE) AMIE+ (YAGO) Inference NELL DeepDive (Tuffy)

slide-22
SLIDE 22

DeepDive (Tuffy)

Rules 1.40 LiveIn(x, y) ← BornIn(x, y) 0.52 BornIn(x, y) ← BornIn(x, z), LocatedIn(z, y) Evidence BornIn(Ruth, Brooklyn) LocatedIn(Brooklyn, NYC) Query LiveIn(?, ?)

slide-23
SLIDE 23

Tuffy Relational Model

BornIn(x, y) ← BornIn(x, z), LocatedIn(z, y)

slide-24
SLIDE 24

Tuffy Relational Model

BornIn(x, y) ← BornIn(x, z), LocatedIn(z, y) SELECT x, y

slide-25
SLIDE 25

Tuffy Relational Model

BornIn(x, y) ← BornIn(x, z), LocatedIn(z, y) SELECT x, y FROM BornIn B, LocatedIn L

slide-26
SLIDE 26

Tuffy Relational Model

BornIn(x, y) ← BornIn(x, z), LocatedIn(z, y) SELECT x, y FROM BornIn B, LocatedIn L WHERE B.y = L.x

slide-27
SLIDE 27

Tuffy Effect

Relational
 Classification Entity Resolution Alchemy Tuffy Alchemy Tuffy Runtime/min 125 250 375 500

3 1 420 68

slide-28
SLIDE 28

Tuffy Limitations

1M Rules 500K Rules 200K Rules 100K Rules Runtime/s 4500 9000 13500 18000

2,196 4,271 9,045 16,507

Tuffy

slide-29
SLIDE 29

The Scalability Challenge

Freebase DBpedia YAGO2s YAGO2

948,047 4,484,907 11,020,000 388,474,630

Publish first Freebase rule set with 36,625 inference rules. Inferred 927M new facts.

slide-30
SLIDE 30

Ontological Pathfinding (OP)

slide-31
SLIDE 31

OP Input

Predicate Subject Object exports United States Computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States

slide-32
SLIDE 32

OP Output

Predicate Subject Object exports Canada Aluminum imports United States Aluminum dealsWith Canada United States

dealsWith(x, y) ← exports(x, z), imports(y, z)

slide-33
SLIDE 33

Experiments

Freebase: largest public knowledge base with 388M facts. Mined 36,625 rules in 33.22 hours. First Freebase rule set.

slide-34
SLIDE 34

Experiments

YAGO YAGO2s Freebase Runtime/min 2000 4000 6000 8000

7,200 293.4 4.56 1,993.2 19.4 3.59

OP AMIE+

slide-35
SLIDE 35

Knowledge Expansion

slide-36
SLIDE 36

Relational Joins Partition- ing Parallel Inference Cross Validation

Knowledge Expansion

slide-37
SLIDE 37

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ

slide-38
SLIDE 38

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ

slide-39
SLIDE 39

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ

slide-40
SLIDE 40

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ

slide-41
SLIDE 41

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ

slide-42
SLIDE 42

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ dealsWith (Canada, United States)

slide-43
SLIDE 43

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M Γ

slide-44
SLIDE 44

Relational Joins

Mining rules requires M ⨝ Γ ⨝ Γ. Worst case complexity |M||Γ|2.

slide-45
SLIDE 45

Relational Joins Partition- ing Parallel Inference Cross Validation

Knowledge Expansion

slide-46
SLIDE 46

“Pick battles big enough to matter, small enough to win.”

— Jonathan Kozol

slide-47
SLIDE 47

Partitioning

Given s and m, can we require each partition: |Γ| <= s, and |M| <= m?

slide-48
SLIDE 48

Partitioning

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M1 Γ1

slide-49
SLIDE 49

Partitioning

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M2 Γ2

slide-50
SLIDE 50

Partitioning

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M1 Γ1 |ΔΓ1| = 2

slide-51
SLIDE 51

Partitioning

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M2 Γ2 |ΔΓ2| = 0 |ΔΓ1| = 2

slide-52
SLIDE 52

Recursive Partitioning

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M1 Γ1

slide-53
SLIDE 53

Recursive Partitioning

p x y exports United States computer exports Canada Aluminum imports United States Aluminum imports United States Clothing dealsWith Canada United States isLocatedIn Washington, D.C. United States isLocatedIn Ottawa Canada isLocatedIn Stanford University Stanford, CA hasCapital Canada Ottawa hasCapital Unites States Washington, D.C. wasBornIn Donald Knuth Milwaukee, Wisconsin isCitizenOf Donald Knuth United States worksAt Donald Knuth Stanford University H(x, y) b1(x, z) b2(y, z) dealsWith exports imports dealsWith isLocatedIn isLocatedIn isCitizenOf wasBornIn HasCapital worksAt wasBornIn isLocatedIn isLocatedIn hasCapital isLocatedIn

M3,4 Γ3,4 Overlapping Independent

slide-54
SLIDE 54

Partitioning

Partitioned joins: M1 ⨝ Γ1 ⨝ Γ1 and M2 ⨝ Γ2 ⨝ Γ2, etc. Worst case complexity |M||γ|2, where γ is the maximum partition size.

slide-55
SLIDE 55

Relational Joins Partition- ing Parallel Inference Cross Validation

Knowledge Expansion

slide-56
SLIDE 56

Parallel Inference

H(x, y) ← b1(x, z), b2(y, z)

slide-57
SLIDE 57

Parallel Inference

F1 F4 F5 R1 R2 R3 F1 R1 R2 R3 F4 R1 R2 R3 F5 R1 R2 R3

slide-58
SLIDE 58

Parallel Inference

F1 F1 R1 F2, F5 R2 F4 R3 F4 F2, F3 R1 F1 R2 F2 R3 F5 F5 R1 F3, F4 R2 F4 R3 F1 0, R1, R2 F2 R1, R2, R3 F3 R1, R2 F4 0, R2, R3 F5 0, R1, R2, R3

slide-59
SLIDE 59

Parallel Inference

F2 F3

slide-60
SLIDE 60

Performance

927M new facts in 19 hours. First inference engine on Freebase.

slide-61
SLIDE 61

Efficiency Improvement

100K Rules 200K Rules 500K Rules 1M Rules Runtime/s 4500 9000 13500 18000

16,507 9,045 4,271 2,196

210 65 30 25

ProbKB Tuffy

slide-62
SLIDE 62

Effect of Partitioning

200M 100M 50M 20M 10M 5M Runtime/h 12.5 25 37.5 50

6.374 9.668 11.842 16.499 24.693 41.915

slide-63
SLIDE 63

Relational Joins Partition- ing Parallel Inference Cross Validation

Knowledge Expansion

slide-64
SLIDE 64

Rules

AMIE+ Validation

YAGO2 YAGO2s

slide-65
SLIDE 65

Facts

AMIE+ Validation

YAGO2 YAGO2s Rules

slide-66
SLIDE 66

AMIE+ Validation

YAGO2 YAGO2s Rules Facts

slide-67
SLIDE 67

AMIE+ Validation

YAGO2 YAGO2s Rules Facts

slide-68
SLIDE 68

Limitations

KB availability. Inference biase.

slide-69
SLIDE 69

Cross Validation

Test Train OP Rules

slide-70
SLIDE 70

Cross Validation

Test Train OP Infer Rules Facts

slide-71
SLIDE 71

Cross Validation

Test Train OP Infer Verify Rules Facts

slide-72
SLIDE 72

Cross Validation

: Inferred facts sorted by descending confidence. Recall: Precision: Recall(Γ+|Γ) = |Γ+ − Γ| |Γ| Γ+ Precision(Γ+|Γ) = |Γ+ ∩ Γ| |Γ+|

slide-73
SLIDE 73

Cross Validation

Precision 0.25 0.5 0.75 1 Recall 0.25 0.5 0.75 1

Freebase YAGO2s

slide-74
SLIDE 74

Inferred Facts

Music Book Film People Location # Correct Inferred Facts 7500000 15000000 22500000 30000000

916,438 1,354,632 1,361,939 1,384,209 21,463,725

slide-75
SLIDE 75

Examples

music/album/artist(Live Era ’87-’93, Guns N’ Roses) book/series_editor/book_edition_series_edited(
 Janet Morris, Heroes in Hell by Baen Books) film/film/production_companies(Butt Spanking, Bacchus) user/anjackson/default_domain/bitstream_encoding/ format(PDF 1.4, Portable Document Format)

slide-76
SLIDE 76

Inferred Errors

bornIn(Mandel, Berlin) bornIn(Mandel, Baltimore) isLocatedIn(Baltimore, Berlin)

slide-77
SLIDE 77

Inferred Errors

isLocatedIn(Baltimore, Berlin) bornIn(Freud, Baltimore) bornIn(Freud, Berlin)

slide-78
SLIDE 78

Functional Constraints

wasBornIn isCitizenOf isMarriedTo isCapitalOf isLocatedIn headquaterIn

slide-79
SLIDE 79

Functional Constraints

bornIn(Mandel, Berlin) bornIn(Mandel, Baltimore)

slide-80
SLIDE 80

Functional Constraints

Precision 0.225 0.45 0.675 0.9 Estimated # of correct facts 7500 15000 22500 30000

Without constraints With constraints

slide-81
SLIDE 81

Error Analysis

Others 3% Incorrect facts 6% Incorrect rules 33% Ambiguous join keys 24% Ambiguities (detected) 34%

slide-82
SLIDE 82

Error Analysis

Others 3% Incorrect facts 6% Incorrect rules 33% Ambiguous join keys 24% Ambiguities (detected) 34%

slide-83
SLIDE 83

Knowledge Activation

slide-84
SLIDE 84

Knowledge Activation

Francis Bacon Aristotle Plato Cicero Philosophy John Locke Meta physics

slide-85
SLIDE 85

Knowledge Activation

Francis Bacon Aristotle Plato Cicero Philosophy John Locke Meta physics i n fl u e n c e d B y

slide-86
SLIDE 86

Knowledge Activation

Francis Bacon Aristotle Plato Cicero Philosophy John Locke Meta physics

mainInterest

slide-87
SLIDE 87

Knowledge Activation

Francis Bacon Aristotle Plato Cicero Philosophy John Locke Meta physics

1.00

slide-88
SLIDE 88

Knowledge Activation

Francis Bacon Aristotle Plato Cicero Philosophy John Locke Meta physics i n fl u e n c e d B y

influencedBy

influencedBy

influence

1.00 0.42 0.29 0.00 0.00

slide-89
SLIDE 89

Knowledge Activation

Francis Bacon Aristotle Plato Cicero Philosophy John Locke Meta physics

1.00 0.42 0.29 0.00 0.00 0.00 0.00

mainInterest

mainInterest

coreSubject

mainInterest

slide-90
SLIDE 90

p x y influencedBy Francis Aristotle influencedBy Francis Plato influencedBy Francis Cicero influence Francis John Locke mainInterest Aristotle Philosophy mainInterest Plato Philosophy coreSubject Cicero Philosophy mainInterest John Locke Meta physics x w Francis 1.0

Q Γ

slide-91
SLIDE 91

p x y influencedBy Francis Aristotle influencedBy Francis Plato influencedBy Francis Cicero influence Francis John Locke mainInterest Aristotle Philosophy mainInterest Plato Philosophy coreSubject Cicero Philosophy mainInterest John Locke Meta physics x w Francis 1.0

Q Γ

x w Francis 1 Francis 3 Plato 3 Aristotle 6

H

slide-92
SLIDE 92

p x y influencedBy Francis Aristotle influencedBy Francis Plato influencedBy Francis Cicero influence Francis John Locke mainInterest Aristotle Philosophy mainInterest Plato Philosophy coreSubject Cicero Philosophy mainInterest John Locke Meta physics x w Francis 1.0

Q Γ

x w Francis 1 Francis 3 Plato 3 Aristotle 6

H

slide-93
SLIDE 93

Query Optimization

Build materialized views; Query assumed to be small; Q ⨝ Γ is efficient with indexes; Use (Q ⨝ Γ) ⨝ H to reduce result size.

slide-94
SLIDE 94

Experiments

Runtime/ms 17.5 35 52.5 70

  • Iter. 1
  • Iter. 2

Iter 3. Query 1Query 2Query 3 Query 1Query 2Query 3 Query 1Query 2Query 3

slide-95
SLIDE 95

Experiments

Runtime/ms 1 100 10000 SemMemDB Douglass

10,900 11.29

More than 500 times of speedup

slide-96
SLIDE 96

Conclusion

slide-97
SLIDE 97

Conclusion

We tackle the knowledge expansion problem. We propose the Ontological Pathfinding (OP) algorithm for scalable rule mining. We extend the OP algorithm for knowledge expansion to infer missing facts in existing knowledge bases. Develop the first mining and inference engine for Freebase.

slide-98
SLIDE 98

Future Work

Extending rule mining to constraint mining. Online and incremental learning over dynamic knowledge bases. Abductive reasoning for query processing.

slide-99
SLIDE 99

Publications (In Submission)

  • 1. Archimedes: Efficient Query Processing over

Probabilistic Knowledge Bases
 Yang Chen, Xiaofeng Zhou, Kun Li, Daisy Zhe Wang
 The SIGMOD Record, 2017

  • 2. Quality Control in Uncertain Knowledge Bases


Daisy Zhe Wang, Yang Chen, Sean Goldberg, Miguel Rodríguez, Yang Peng
 20th International Conference on Extending Database Technology, 2017

slide-100
SLIDE 100

Publications

  • 3. ScaLeKB: Scalable Learning and Inference over Large

Knowledge Bases
 Yang Chen, Daisy Zhe Wang, Sean Goldberg
 The VLDB Journal, 2016

  • 4. ArchimedesOne: Query Processing over Probabilistic

Knowledge Bases
 Xiaofeng Zhou, Yang Chen, Daisy Zhe Wang
 Proceedings of the VLDB Endowment, 2016

slide-101
SLIDE 101

Publications

  • 5. Ontological Pathfinding: Mining First-Order Knowledge

from Large Knowledge Bases
 Yang Chen, Sean Goldberg, Daisy Zhe Wang, Soumitra Siddharth Johri
 Proceedings of the ACM SIGMOD International Conference on Management of Data, 2016

  • 6. Efficient In-Database Analytics with Graphical Models


Daisy Zhe Wang, Yang Chen, Christan Grant, Kun Li
 IEEE Data Engineering Bulletin, 2014

slide-102
SLIDE 102

Publications

  • 7. Knowledge Expansion over Probabilistic Knowledge

Bases
 Yang Chen, Daisy Zhe Wang
 Proceedings of the ACM SIGMOD International Conference on Management of Data, 2014

  • 8. SemMemDB: In-Database Knowledge Activation


Yang Chen, Milenko Petrovic, Micah H. Clark
 Proceedings of the Twenty-Seventh International Florida Artificial Intelligence Research Society Conference (FLAIRS), 2014

slide-103
SLIDE 103

Publications

  • 9. Web-Scale Knowledge Inference Using Markov Logic Networks


Yang Chen, Daisy Zhe Wang
 Proceedings of ICML workshop on Structured Learning: Inferring Graphs from Structured and Unstructured Inputs (SLG), 2013

  • 10. Automatic Knowledge Base Construction Using Probabilistic

Extraction, Deductive Reasoning, and Human Feedback
 Daisy Zhe Wang, Yang Chen, Sean Goldberg, Christan Grant, Kun Li
 Proceedings of the Joint Workshop on Automatic Knowledge Base Construction and Web-scale Knowledge Extraction (AKBC-WEKEX), 2012

slide-104
SLIDE 104

Acknowledgement

  • Dr. Daisy Zhe Wang
  • Dr. Alin Dobra
  • Dr. Anand Rangarajan
  • Dr. Jih-Kwon Peir
  • Dr. Kshitij Khare

Miguel E. Rodrguez Xiaofeng Zhou Dihong Gong Ali Sadeghian Mebin Jacob

  • Dr. Christan Grant
  • Dr. Kun Li
  • Dr. Morteza Shahriari Nia

Sean Goldberg Yang Peng

slide-105
SLIDE 105

Thank you!

Open source:
 http://dsr.cise.ufl.edu/projects/probkb-web- scale-probabilistic-knowledge-base. Questions?

Data Science Research

@