Overview

• Separate instances of ElasticSearch
  • Topic Map
    • Serves as a symbol library for all topics
  • IDocument Collection
    • IWordGram Collection
      • Serves as a neural-like network of all harvested text resources
  • Conceptual Graph
    • Serves as a knowledge model to tie together all topics
  • TSC
    • Qualitative process models derived from harvesting
• Anticipatory Behavior

Big Picture Diagram

  

Topic Map

• Well-merged Graph
• Created through
  • Ontology import
  • Machine reading
  • Social interaction
    • Knowledge Gardening

IDocument Collection

• An IDocument is a container which shepherds all harvesting activities related to
  • Any document harvested through machine reading
  • Any Ontology Class which has descriptions
  • Any Topic Node which has descriptions

IWordGram Collection

• Their place in the hierarchy of concepts
  • IDocument representation of harvested document
    • IParagraph collections
      • ISentence Collections
        • SentenceScanner
          • Uniform text processor
        • IWordGram collections
          • All IWordGram collections are pooled across all harvested documents
            • No duplicate objects: one serves all instances
          • ITuple Collections
            • The final workhorses used by
              • Topic Map
              • Conceptual Graphs
              • TSC

 TSC

• Qualitative Process Theory [Ken Forbus]
  • Actors
  • Relations
  • States
  • Process Rules
  • Episodes
    • Process Models
• Created while harvesting
• Created and updated in background tasks
• Used in a case-based-reasoning way while harvesting
  • Provides anticipatory clues

Conceptual Graph

• Components [John Sowa]
  • Concept Lattice
  • Predicate Lattice
  • Graphs
    • Knowledge Graphs
      • Harvested
    • Canonical Graphs
      • Mapped
        • Category Theory?

Anticipatory Behavior

• Means by which harvesting occurs
  • What is already known helps form expectations of what is being harvested next
    • Qualitative models form expectations
    • Canonical graphs are anticipatory schemas for knowledge modeling
    • Topic Map helps keep named entities and relations well organized

First-Order Workflow

• Pretrain
  • Import Ontologies, etc
• Read text
  • New IDocument per document read
  • Scan for paragraphs (IParagraph)
  • For each paragraph
    • Scan for sentences (ISentence)
      • For each sentence
        • Scan for WordGrams (IWordGram)
        • Scan for Tuples (ITuple)
      • Sentences without Tuples are un-resolved sentences
• Second-Order Workflow

Second-Order Workflow

• Model Building
  • Conceptual Graphs
  • Process Models
  • Probability Models
• Use Models to
  • Help solve un-resolved sentences
  • Help answer questions
  • Support Third-Order Workflow
• Generate research tasks to further solve un-resolved sentences
  • Web search
  • User interaction
• Third-Order Workflow

Third-Order Workflow

• Generalized background tasks
  • Wide-ranging ​topic map merge ​studies
  • Wide-ranging ​contradiction​ studies
  • Wide-ranging ​un-connected dot​ studies
  • Wide-ranging ​link validation​ studies