210 likes | 366 Views
Semantic Network Complexity And Metathesaurus Maintenance. Stuart Nelson, MD “Dr. Ad Hoc”. Role of Semantic Network. Categorize the Concepts of the Metathesaurus Model Reality Model the Contents. Semantic Network Problems. Structural Considerations Limited Use Poor Type Assignments.
E N D
Semantic Network ComplexityAnd Metathesaurus Maintenance Stuart Nelson, MD “Dr. Ad Hoc”
Role of Semantic Network • Categorize the Concepts of the Metathesaurus • Model Reality • Model the Contents
Semantic Network Problems • Structural Considerations • Limited Use • Poor Type Assignments
Problematic Assignments • Inconsistent Assignments • “Wrong” Assignments • Strange Type Co-Occurrences
Why Do These Things Happen” • Stupid Editors? • Poor Selection of Types? • Inherent Complexity? • Poor Management
Where Does Complexity Arise, Or Why Is This So Hard? • Formalization Gap • Duality of Central Notions • Boundary Phenomenon • Nelson’s Lemma
Formalization Gap Natural Language Understanding Pattern Recognition Recall and Reliable Memory Formal Reasoning Formalization Gap
A Different Aspect • Humans Have Difficulty With Formal Notions • Formal Definitions • A set of ordered pairs • The set of first coordinates of the ordered pairs
Duality of Central Notions • Cancer – Process or Object? • Gene – Locus or Sequence? • Finding - Event or Relationship? • Waves and Particles
Boundary Phenomenon • Definitional Characteristics • Necessary and Sufficient Attributes
Nelson’s Lemma • Greater Number of Boundaries LEADS TO • Greater Number of Borderline Cases • Overlapping “Extents”
Process of Semantic Type Assignment • Default Assignment from Source • Some Sources (MeSH) have STYs • Human Review of Concept Status • Human Review of Semantic Types
Practical Considerations • Achieving Consistency • Managing Changes • Distribution of STY Assignments • Understanding Source Meanings • URU Criteria
Achieving Consistency • See Nelson’s Lemma • Internal (Self) Consistency • Editorial Variability • Inconsistent Rules • Use Parent • Use Several Children
Managing Changes • How Many Seconds per Change? • How Many Records to Edit? • What Can Be Done Algorithmically? • Resources Available
Distribution of Assignments • Overlapping Meanings • Facet-Based Typing • Inconsistent Rules
Understanding Source Meanings • Definitions Rare • “Contextual Understanding” • Hierarchies • Most Vocabularies A Set of Engineering Compromises
Prime Directive • Respect the Meaning in the Source • Names Lack Face Validity • Reification of Relationships
Respect • Reality - Names and Rules Only Reflect Our Understanding • Sources - Reflect The View of The Source
URU Criteria For Types • Understandable • Reproducible • Usable
Challenges • Analysis • Useful algorithms • Time Involved • Assessing Benefit • Training • Establishing Business Rules • Testing Consistency