Graph Gurus Episode 14: Pattern Matching

Graph Gurus
Episode 14
Pattern Matching using GSQL Interpreted Mode

© 2019 TigerGraph. All Rights Reserved
Welcome
● Attendees are muted
● If you have any Zoom issues please contact the panelists via chat
● We will have 10 min for Q&A at the end so please send your questions
at any time using the Q&A tab in the Zoom menu
● The webinar will be recorded and sent via email
2

Today's Gurus
3
Victor Lee
Director, Product Management
● BS in Electrical Engineering and Computer
Science from UC Berkeley
● MS in Electrical Engineering from Stanford
University
● PhD in Computer Science from Kent State
University focused on graph data mining
● 15+ years in tech industry
Mingxi Wu
VP, Engineering
● BS in Computer Science from Fudan University,
China
● MS in Computer Science from University of Florida
● PhD in Computer Science from University of
Florida
● 19+ years in data management industry &
research

Agenda
4
● Pattern Matching Uses in Data Analytics
● GSQL Pattern Matching Syntax and Semantics
● Pattern Matching Using LDBC SNB Data (Demo)
● A Recommendation Application (Demo)
● Q&A

Graph Use Cases
5
By Dan McCreary https://medium.com/@dmccreary/a-taxonomy-of-graph-use-cases-2ba34618cf78

Graph Use Cases
6
Rule
Enforcement
Flexible
Modeling, Data
Integration
Insight from
Relationships
Reporting,
Auditing
Clustering,
Community,
Similarity,
Ranking
Finding
Connections,
Deep Link
Analysis
Anti-Fraud
Recommendation
Master Data
Management
Machine Learning,
Explainable AI
Entity
Resolution
Ad Hoc
Exploration,
Visual
Explainable
Results
Knowledge
Inference
Risk
Management
Data
Provenance

Graph Use Cases
7
Rule
Enforcement
Flexible
Modeling, Data
Integration
Insight from
Relationships
Reporting,
Auditing
Clustering,
Community,
Similarity,
Ranking
Finding
Connections,
Deep Link
Analysis
Anti-Fraud
Recommendation
Master Data
Management
Machine Learning,
Explainable AI
Entity
Resolution
Ad Hoc
Exploration,
Visual
Explainable
Results
Knowledge
Inference
Risk
Management
Data
Provenance
Pattern
Matching/Search

Graph Patterns
Any arrangement of connected nodes, used as a search input:
"Find all the occurrences like this"
Pattern Types
● Transactional Patterns
● Social Patterns
● Event Patterns
● Hybrid Patterns
Uses
● Recommendation
● Fraud/Crime
● Security/Risk/Rules
● Trends/Insight

© 2019 TigerGraph. All Rights Reserved 9
Detecting Phone-Based Fraud by Analyzing Network or Graph
Relationship Features at China Mobile
Download the solution brief at - https://info.tigergraph.com/MachineLearning

Pattern Matching Overview
10
● What is a pattern?
○ Pattern is a traversal trace on the graph schema
○ Use regular expression to represent the repetitive steps
○ Pattern can be linear, or nonlinear (tree, circle etc.)
● Example
○ Imagine a schema consisting of a Person vertex type and a
Friendship edge type.
○ A pattern
■ Person-(Friendship)-Person-(Friendship)-Person
■ Person-(Friendship*2)-Person

Pattern Matching Overview
11
● What is Pattern Matching?
○ Pattern matching is the process of finding subgraphs in a
data graph that conforms to a given query pattern.
○ Input
■ Query Pattern P
■ Data Graph G
○ Output
■ Collection of matched instances M = {subgraphs of G}

Query Pattern P
Data Graph G

Query Pattern P
Data Graph G
First Match

Query Pattern P
Data Graph G
First Match
Second Match

Query Pattern P
Data Graph G
First Match
Second Match
Third Match

GSQL Pattern Matching Overview
16
● Simple 1-Hop Pattern
● Repeating a 1-Hop Pattern
● Multi-Hop Linear Pattern

Simple 1-Hop Pattern
17
Classic GSQL Syntax:
FROM X:x -((E1 | E2 | E3):e1)-> Y:y
● Semantics: Traverse from Left to Right
GSQL Pattern Matching Syntax:
FROM X:x -((E1> | <E2 | E3):e1)- Y:y
● Semantics: Pattern is a whole;
each edge's direction is a detail about the pattern

Simple 1-Hop Pattern - Variations
18
1. FROM X:x -(E1:e1)- Y:y
○ E1 is an undirected edge. x, y bind to the end points of E1. e1 is alias to E1.
2. FROM X:x -(E2>:e2)- Y:y
○ Right directed edge, x binds to the source of E2, y binds to the target of E2. e2 is alias of E2.
3. FROM X:x -(<E3:e3)- Y:y
○ Left directed edge, y binds to the source of E3, x binds to the target of E3. e3 is alias of E3.
4. FROM X:x -(_:e)- Y:y
○ Any undirected edge pattern. "_" means any undirected edge type. e is alias of "_".
5. FROM X:x -(_>:e)- Y:y
○ Any right directed edge. "_>" means any right directed edge type. e is alias of "_>".

Simple 1-Hop Pattern - Variations
19
6. FROM X:x -(<_:e)- Y:y
○ Any left directed edge. "<_" means any left directed edge type. e is alias of "<_".
7. FROM X:x -((<_|_):e)- Y:y
○ Any left directed or any undirected. "|" means OR. e is alias of (<_|_).
8. FROM X:x -((E1|E2>|<E3):e) - Y:y
○ Disjunctive 1-hop edge. (x,y) are connected by E1 or E2 or E3. e is alias of (E1|E2>|<E3).
9. FROM X:x -()- Y:y
○ any edge (directed or undirected) match this 1-hop pattern
○ Same as this: (<_|_>|_)

Simple 1-Hop Pattern (cont.)
20

Repeating a 1-Hop Pattern - Unbounded Variations
22
1. FROM X:x - (E1*) - Y:y
○ X connect to Y via 0 or more repetition of undirected E1. * is called the Kleene star.
2. FROM X:x - (E2>*) - Y:y
○ X connect to Y via 0 or more repetition of rightward E2
3. FROM X:x - (<E3*) - Y:y
○ X connect to Y via 0 or more repetition of leftward E3
4. FROM X:x - (_*) - Y:y
○ X connect to Y via 0 or more undirected edges of any type

Repeating a 1-Hop Pattern - Unbounded Variations
23
5. FROM X:x - (_>*) - Y:y
○ X connect to Y via 0 or more right-directed edges of any type
6. FROM X:x - (<_*) - Y:y
○ X connect to Y via 0 or more left-directed edges of any type
7. FROM X:x - ((E1|E2>|<E3)*) - Y:y
○ Either E1, E2> or <E3 can be chosen at each repetition.

Repeating a 1-Hop Pattern - Bounded Variations
24
1. FROM X:x - (E1*2..) - Y:y
○ Lower bounds only. There is a chain of at least 2 E1 edges.
2. FROM X:x - (E2>*..3) - Y:y
○ Upper bounds only. There is a chain of between 0 and 3 E2 edges.
3. FROM X:x - (<E3*3..5) - Y:y
○ Both Lower and Upper bounds. There is a chain of 3 to 5 E3 edges.
4. FROM X:x - ((E1|E2>|<E3)*3) - Y:y
○ Exact bound. There is a chain of exactly 3 edges, where each edge is either E1, E2>, or <E3.

Repeating a Pattern - Remarks
25
1. Edge aliases cannot be used together with a Kleene star.
○ When an edge is repeated, it would not clear which instance of the edge
the alias refers to.
2. Shortest path semantics.
○ When an edge is repeated with the Kleene star, we consider shortest path
matches only.
○ Counting and existence check of matched shortest path pattern is
tractable.

Example: Shortest Matching Path
In above figure, for the pattern, 1 - (E>*) - 4, the following paths can reach 4 from 1.
1. 1->2->3->4
2. 1->2->3->5->6->2->3->4 (1 loop)
3. Any path 1>2, looping through 2->3->5->6->2 two or more times, and exiting
2>3>4
Only the first one is the shortest, and considered a match for the pattern.

Specifying How Many Repetitions (cont.)

Multiple-Hop Linear Pattern
29
● 2-hop pattern
○ FROM X:x - (E1:e1) - Y:y - (E2>:e2)-Z:z
1st hop 2nd hop
Similarly, a 3-hop pattern concatenates three 1-hop patterns in sequence,
every two connecting patterns share one endpoint. Below Y:y and Z:z are
the connecting endpoints.
● 3-hop pattern
○ FROM X:x - (E2>:e2) - Y:y - (<E3:e3)- Z:z - (E4:e4) - U:u
1st hop 2nd hop 3rd hop

Multiple-Hop Linear Pattern
30
● SELECT Clause
○ select pattern’s endpoints only
○ FROM X:x-(E2>:e2)-Y:y-(<E3:e3)-Z:z-(E4:e4)-U:u;
● FROM Clause
○ FROM pattern can be shortened
■ FROM X:x-(E2>:e2)-Y:y-(<E3:e3)-Z:z-(E4:e4)-U:u ⇒
■ FROM X:x-(E2>.<E3.E4)-U:u; //note: no alias is allowed for E.E style
● WHERE Clause
○ Conjunction of local hop predicate only
○ Last hop local predicate can refer to the starting end point
○ Kleene star breaks local hop predicate
● ACCUM and POST-ACCUM Clause
○ ACCUM clause executed for each matched instance
○ ACCUM to endpoints only

How To Use In Upcoming Release 2.4
31
● Session Parameter
○ set syntax_version="v2" //for pattern match
○ set syntax_version="v1" //default, classic syntax
● Invoke Interpret GSQL Engine
○ INTERPRET QUERY QueryName (p1, p2, ..) //run existing query
○ INTERPRET QUERY () FOR GRAPH graphName { } //adhoc query
● Query Level Syntax Setting
○ CREATE QUERY (p1, p2.,.) Name FOR GRAPH test SYNTAX("v2") { }
○ INTERPRET QUERY () FOR GRAPH test SYNTAX("v1") { }

Demo Setup
32

Demo Setup
33
E1 1,003,605
E2 2,052,169
E3 1,003,605
E4 229,166
E5 1,611,869
E6 90,492
E7 2,698,393
E8 713,258
E9 309,766
E10 16,080
E11 1,575
E12 6,380
E13 2,052,169
E14 1,003,605
E15 9,892
E16 1,343
E17 111
E18 70
Total :
3.18M vertices
17.25M edges
11 vertex types
25 edge types
Vertex Stats
Edge Stats
Total comment post
compan
y university city country continent forum person tag tagclass
# attributes 6 8 3 3 3 3 3 3 10 3 3
# vertices 3.18 M 2,052,169 1,003,605 1,575 6,380 1,343 111 6 90,492 9,892 16,080 71
E19 180,623
E20 1,438,418
E21 751,677
E22 1,040,749
E23 1,011,420
E24 7,949
E25 21,654

Demo Setup
34

Demo On Docker
35

Summary
36
Version 2.4
Available End Jun
● Pattern Matching enables
● Demo: Pattern Match Using LDBC Social Network Benchmark Data
○ Interpreted Mode on laptop Docker -- run queries instantly as
soon as you write them
● Demo: Pattern Matching in a Recommender Application
○ Interpreted Mode again

Q&A
Please send your questions via the Q&A menu in Zoom
37

Developer Edition Available
We now offer Docker versions and VirtualBox versions of the TigerGraph
Developer Edition, so you can now run on
● MacOS
● Windows 10
● Linux
Developer Edition Download https://www.tigergraph.com/developer/
38

NEW! Graph Gurus Developer Office Hours
39
Catch up on previous episodes of Graph Gurus:
https://www.tigergraph.com/webinars-and-events/
Every Thursday at 11:00 am Pacific
Talk directly with our engineers every
week. During office hours, you get
answers to any questions pertaining to
graph modeling and GSQL
programming.
https://info.tigergraph.com/officehours

Additional Resources
40
New Developer Portal
https://www.tigergraph.com/developers/
Download the Developer Edition or Enterprise Free Trial
https://www.tigergraph.com/download/
Guru Scripts
https://github.com/tigergraph/ecosys/tree/master/guru_scripts
Join our Developer Forum
https://groups.google.com/a/opengsql.org/forum/#!forum/gsql-users
@TigerGraphDB youtube.com/tigergraph facebook.com/TigerGraphDB linkedin.com/company/TigerGraph

Graph Gurus Episode 14: Pattern Matching

More Related Content

What's hot (20)

Similar to Graph Gurus Episode 14: Pattern Matching (20)

More from TigerGraph (20)

Recently uploaded (20)

Graph Gurus Episode 14: Pattern Matching