Maverick: Discovering Exceptional Facts from Knowledge Graphs - - PowerPoint PPT Presentation

maverick discovering exceptional facts from knowledge
SMART_READER_LITE
LIVE PREVIEW

Maverick: Discovering Exceptional Facts from Knowledge Graphs - - PowerPoint PPT Presentation

Maverick: Discovering Exceptional Facts from Knowledge Graphs 12/03/19 Paper published in Proc. ACM SIGMOD International Conference on Management of Data, 2018. Presented by: Juan Carrillo Candidate for MASc. in Computer Software Department


slide-1
SLIDE 1

Maverick: Discovering Exceptional Facts from Knowledge Graphs

Paper published in Proc. ACM SIGMOD International Conference on Management of Data, 2018. Presented by: Juan Carrillo Candidate for MASc. in Computer Software Department of Electrical & Computer Engineering University of Waterloo

12/03/19

slide-2
SLIDE 2

Agenda

  • 1. Introduction
  • 2. Maverick core features
  • 3. Experiments
  • 4. Conclusions
  • 5. Discussion

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 2

slide-3
SLIDE 3

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 3

Introduction

1

slide-4
SLIDE 4
  • 1. Introduction

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 4

From knowledge graphs to exceptional facts

slide-5
SLIDE 5
  • 1. Introduction

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 5

Knowledge graphs (Linked Data) Maverick approach

Manually designed queries Automated detection of exceptional facts

The problem, and the Maverick approach

Pattern generator

Context evaluator Exceptionality evaluator

Fact reporter

slide-6
SLIDE 6
  • 1. Introduction

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 6

Related background

2010

Outlier detection J Gao - 2010

On community

  • utliers and their

efficient detection in information networks 2016

Outlying aspect mining F Angiulli - 2016

Outlying property detection with numerical attributes 2018

Maverick: Discovering Exceptional Facts from Knowledge Graphs SIGMOD’18

Comprehensive description and math basis 2018

VLDB’18

High-level description and demo

slide-7
SLIDE 7

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 7

Maverick core features

2

slide-8
SLIDE 8
  • 2. Maverick core features

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 8

Entity, context, pattern

slide-9
SLIDE 9
  • 2. Maverick core features

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 9

The overall framework

slide-10
SLIDE 10
  • 2. Maverick core features

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 10

Main Algorithm

slide-11
SLIDE 11
  • 2. Maverick core features

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 11

Description of components

Context Evaluator

  • Uses a graph query

system (Neo4j)

  • Takes a pattern as

input and returns the matches

  • Agnostic to query

processing system

Exceptionality Evaluator Pattern generator

  • Takes the entity of

interest and its contexts

  • Looks for the k

subspaces with highest scores

  • Implements scoring

functions

  • Uses beam search to

look for promising patterns

  • Implements domain

specific heuristics

  • Beam width can be

tuned to requirements

slide-12
SLIDE 12

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 12

Experiments

3

slide-13
SLIDE 13
  • 3. Experiments

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 13

Experimental setup

Single node: 16-core, 32GB RAM

Datasets

WCGoals 49.078 nodes, 158.114 edges, 13 different edge labels, and 11 entity types. OscarWinners 42.148 nodes, 63.187 edges, 24 distinct edge labels, and 13 entity types.

Methods compared

▪ Beam-Rdm ▪ Beam-Opt ▪ Beam-Conv ▪ Breadth-First

slide-14
SLIDE 14
  • 3. Experiments

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 14

Efficiency

slide-15
SLIDE 15
  • 3. Experiments

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 15

Efficiency

slide-16
SLIDE 16
  • 3. Experiments

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 16

Effectiveness

slide-17
SLIDE 17

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 17

Conclusions

4

slide-18
SLIDE 18
  • 4. Conclusions

✓ The authors model an exceptional fact as a context-pattern pair on a

knowledge graph

✓ Exponential complexity of search is handled using beam search ✓ The framework is adaptable to domain specific requirements

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 18

Takeaways and paper contributions

slide-19
SLIDE 19

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 19

Thanks for your attention

slide-20
SLIDE 20

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 20

Discussion

5

slide-21
SLIDE 21
  • 5. Discussion
  • 1. What other heuristics could be proposed in addition to the two presented

in the paper? Design requirements for a third heuristic?

  • 2. How Maverick would perform over a completely different dataset?

Different proportions among nodes, edges, edge labels, and entity types.

  • 3. What if we add attributes to the nodes and edges? Constraints
  • 4. How to adapt Maverick to work over multiple/linked knowledge graphs?

Maverick: Discovering Exceptional Facts from Knowledge Graphs

PAGE 21

Research Industry

  • 5. What is an example of an application over Google knowledge graph?