Learning to Reason Knowledge Acquisition in Cyc
Lecture slides:
- Learning to Reason: Knowledge Acquisition in Cyc
- Overview
- Expressivity
- The Cyc Knowledge Base
- Syntactic Power
- For Inference: Senses of ‘In’
- Senses of ‘In’
- Concepts are densely related
- in Cyc
- Power
- The Cyc Analytic Environment - 1
- The Cyc Analytic Environment - 2
- How is this done?
- The Cyc Analytic Environment - 3
- 45th’s Space Wing Hurricane Preparedness
- Performance: Subtheory: disjointWith
- Inference is Fast & Trainable
- You can get Cyc
- Cycorp Corporate Mission
- Manual Knowledge Entry
- Cycorp Corporate Mission
- Ambitious Approach: General Automated Interview - 1
- Ambitious Approach: General Automated Interview - 2
- Ambitious Approach: Analogical Reasoning
- Ambitious Approach: Using Background Knowledge
- Concept Refinement Interview
- Ambition is Good but...
- What can you do?
- Example: Getting Knowledge in Context
- Example: Use of Acquired Knowledge - 1
- Example: Use of Acquired Knowledge - 2
- Example: Use of Acquired Knowledge - 3
- Example: Use of Acquired Knowledge - 4
- Intelligent Search
- Content adaptation: heart valve repair
- Content adaptation: coronary artery
- Facts and Rules (from 1998, 2003)
- Some opportunities for ML
- Document Tagging
- Military Taxonomy
- Knowledge-based disambiguation - 1
- Disambiguation Rules: ‘Jet’
- Knowledge-based disambiguation - 2
- Knowledge Driven Disambiguation - 1
- Rule Induction
- The Induction Pipeline
- Example - 1
- Example - 2
- Early Results
- Sample Rules Produced
- Situation Recognition
- Example - 1
- Example - 2
- Example - 3
- Example - 4
- Other Approaches
- Previous Results on Whodunit Task
- Integrating Markov Logic
- Early ML Experiments
- Markov Logic Work
- Getting Ground Facts
- Second Life
- Overview
- Cyc
Author: Michael Witbrock, Cycorp Inc.