4. Document Discovery with Graph Data Science

Neo4j, Inc. All rights reserved 2021
Document Discovery with Graph
Data Science
Gary Mann
gary.mann@neo4j.com

● Customers are already very good at search.
○ Too much data, too many tools.
○ Data is highly variable - schemaless.
○ Need to focus analysts.
○ Start with unstructured search - then traverse and discover paths
through data.
● Customers want to:
○ Leverage both structured and unstructured data.
○ Support/discover multiple relationships between pieces of information.
○ Understand how entities interact
○ Navigate/Traverse across both structured and unstructured data
visually.
Graph-Aided Discovery

Graph-Aided Discovery
● Why Graph?
○ Flexible and (mostly) schemaless model
○ Unstructured queries to enter the graph
○ Queries and pattern matching to traverse relationships.
○ Pathfinding algorithms to understand how entities are
related.
○ Graph Analytics to leverage the overall graph structure.
○ Graphs can form the basis for many analytical use-cases
including Discovery, Analytics, Investigations, etc.

Query (e.g. Cypher/Python)
Real-time, local decisioning
and pattern matching
Graph Algorithms
Global analysis
and iterations
You know what you’re
looking for and making a
decision
You’re learning the overall structure
of a network, updating data, and
predicting
Local
Patterns
Global
Computation
When Do I need Graph Algorithms?

Questions to Answer
● Local Patterns
○ How many X are related to Y?
○ How are X and Y related through multiple hops?
○ What characteristics to Y and Z share in common?
■ Are they the same entity?
MATCH p=(a:Person)-[:HAS_PHONE]->(b:Person) RETURN p
● Global Analytics
○ What are my important entities in my graph?
○ What data is related through its relationships in the graph?
○ Can I predict relationships that don’t explicitly exist?

The Neo4j Graph Data Science Library
• Deep path analytics
• Optimal routing
• Evaluates how alike nodes are
• Construct graphs from data
Pathfinding
& Search
Similarity
Community
Detection
Mutable In-Memory
Workspace
Computational Graph
Native Graph Store
50+ Robust Algorithms
Flexible Analytics Workspace
• Identifies node importance
• Influencer & Risk Identification
Centrality /
Importance
• Detects group clustering
• Partition options
• Estimates likelihood of
• Estimate missing information
Link
Prediction
Graph
Embeddings
• Learn your graph topology
• Use for dimensionality reduction

7
Supervised ML
Graph-Native
Feature
Engineering
Train
Predictive
Model
Queries
Algorithms
Embeddings
1. Model Type
2. Property
Selection
3. Train & Test
4. Model
Selection
Apply Model to
Existing / New
Data
Store Model in
Database
Use Predictions
for Decisions
Use Predictions
to Enhance
the Graph
Publish & Share

1. Ingest Data and Derive Relationships from Unstructured pieces
2. Leverage Graph Analytics to Enhance the Data/Discovery Process
3. Enable Unstructured Search on the Graph
4. Enable Visual Exploration through Neo4j Bloom
5. Tie it all together in a simple Web Application
Example Use-Case

9
Custom Application

Demo

● Graphs can provide context to information discovery applications.
○ Data relationships are key
○ Both structured and unstructured data
○ Exploration and analytics
○ Focus analysts along paths
○ Graph Data Science algorithms used to enhance the graph.
● Enhanced Document Graphs can form the basis of many discovery and
analytical uses cases.
Takeaways

12
Thank you!
Contact us at
gary.mann@neo4j.com

4. Document Discovery with Graph Data Science

More Related Content

What's hot

Similar to 4. Document Discovery with Graph Data Science

More from Neo4j

Recently uploaded

4. Document Discovery with Graph Data Science