Knowledge Graphs as a Pillar to AI
0 likes1,303 views
The document discusses the role of knowledge graphs in artificial intelligence, emphasizing their integration with ontologies and graph databases. It presents two use cases: using knowledge graphs as recommendation engines and for high precision auto-tagging, highlighting their importance in organizing knowledge. The document explains the differences and relationships among folksonomies, taxonomies, and ontologies, and showcases examples of practical applications in a business context.
![P & P: ATTRIBUTES AND RELATIONS
People Presentations
Name
Job Title
Employer
[…]
Title
Description
Topics*
Attend(s)
Has Speaker
Spoke At](https://image.slidesharecdn.com/knowledgegraphsasapillartoai-180613134914/85/Knowledge-Graphs-as-a-Pillar-to-AI-10-320.jpg)
![P & P: EXAMPLE GRAPH
CS203
.
.
Knowledge Graphs
as a Pillar to AI
Attend
James Midkiff
Developer
Yanko Ivanov
Senior KM Consultant
Audience Member
[Role]](https://image.slidesharecdn.com/knowledgegraphsasapillartoai-180613134914/85/Knowledge-Graphs-as-a-Pillar-to-AI-11-320.jpg)
Knowledge Graphs as a Pillar to AI
- 1. KNOWLEDGE GRAPHS AS A PILLAR TO AI Yanko Ivanov and James Midkiff May 23, 2018
- 2. TABLE OF CONTENTS ONTOLOGY + GRAPH DATABASE = KNOWLEDGE GRAPH USE CASE 1: KNOWLEDGE GRAPH AS A RECOMMENDATION ENGINE USE CASE 2: HIGH PRECISION AUTO-TAGGING
- 3. TAXONOMIES, ONTOLOGIES, AND KNOWLEDGE GRAPHS AS PART OF AI
- 4. KNOWLEDGE ORGANIZATION CONTINUUM FOLKSONOMY Free-text tags. CONTROLLED LIST List of pre-defined terms. Improves consistency. THESAURUS Pre-defined terms & synonyms. Hierarchical relationships. Associative (“related to”) relationships. Scope notes. Increased expressiveness. TAXONOMY Pre-defined terms & synonyms. Hierarchical relationships. Improves consistency. Allows for parent/child content relationships. ONTOLOGY Scope notes. Pre-defined classes & properties. Expanded relationship types. Increased expressiveness. Semantics. Inference.
- 5. BUSINESS ONTOLOGY A defined data model that describes structured and unstructured information through: • entities, • their properties, • and the way they relate to one another. • Ontology is about things, not strings. • Ontologies model your domain in a machine and human understandable format. • Ontologies provide context. • Effective ontologies require a deep understanding of the knowledge domain.
- 6. GRAPH DATABASE ▪ A linked data store that organizes structured and unstructured information through: ▪ entities, ▪ their properties, ▪ and relationships. ▪ Also known as: ▪ Linked Data Store (LD Store) ▪ Triple Store ▪ “Knowledge Graph” ▪ Consists of triples Subject Predicate Object Project A hasTitle Title A Person B isPMOn Project A Document C isAbout Topic D Document C isAbout Topic F Person B IsExpertIn Topic D … … …
- 7. KNOWLEDGE GRAPH Content Sources Subject Predicate Object Project A hasTitle Title A Person B isPMOn Project A Document C isAbout Topic D Document C isAbout Topic F Person B IsExpertIn Topic D … … … Business Ontology Graph Database Enterprise Knowledge Graph Business Taxonomy Person B Project A Document C Person F Topic D Topic E
- 8. EXAMPLE USE CASE #1: PEOPLE AND PRESENTATIONS
- 9. PEOPLE AND PRESENTATIONS (P & P) People Presentations
- 10. P & P: ATTRIBUTES AND RELATIONS People Presentations Name Job Title Employer […] Title Description Topics* Attend(s) Has Speaker Spoke At
- 11. P & P: EXAMPLE GRAPH CS203 . . Knowledge Graphs as a Pillar to AI Attend James Midkiff Developer Yanko Ivanov Senior KM Consultant Audience Member [Role]
- 12. P & P: OUR PRESENTATION TOPICS Knowledge Graphs Artificial Intelligence Ontology Design Taxonomy Design Machine Learning Graph Technology Recommendation Engine Semantic AI Audience Member Has Interest
- 13. P & P: SIMILARITY BY INTEREST How similar are you to the person next to you? How many interests do you share? How many unique interests do you both have total? Jaccard Index "coefficient de communauté" by Paul Jaccard
- 14. P & P: PROBABILITY FOR TOPICS P(A,T) = Probability that Audience Member A has an interest in Topic T Knowledge Graphs Artificial Intelligence Ontology & Taxonomy Design Machine LearningGraph Technology Recommendation Engine Semantic AI
- 15. P & P: PROBABILITY FOR PRESENTATIONS P(A,P) = Probability that Audience Member A would attend Presentation P CS203 Attend
- 16. P & P: RECOMMENDATION ADD-ONS Weighted Interests • Scale based on • Frequency • Time User Input and/or Feedback • User specifies topics • User liked or disliked recommendation Auto-Tagging Presentations • Manual tagging is inconsistent • Text extraction provides context
- 17. EXAMPLE USE CASE #2: AUTO-TAGGING
- 18. AUTO-TAGGING ▪ Problem (Context) Specific Knowledge Graph ▪ Ontology for Content Tagging ▪ Enables Data Analysis on all Content Taxonomy Content Tag
- 19. AUTO-TAGGING EXTENSIONS ▪ Enhanced Auto-Tagging ▪ History of Documents ▪ Implicit Auto-Tagging ▪ Associate Taxonomy Terms ▪ Classification ▪ Group Content based on Tags Taxonomy Content Tag Co-occurrence