Concept Description Vectors and the 20 Questions Game

1
Concept Description Vectors and the 20 Questions
Game

• Wlodzislaw Duch
• Tomasz Sarnatowicz
• Julian Szymanski

2
Semantic Memory

• Permanent container for general knowledge

3
Hierarchical Model Collins Quillian, 1969
4
Semantic network Collins Loftus, 1975
5
Semantic Memory
6
Semantic Space

• All the concepts and keywords create a
• semantic matrix

7
Concept Description Vectors

• CDV a vector of properties describing a single
  concept
• Most of elements are 0s sparse vector

8
Data Sources I

• Machine readable dictionaries and ontologies
• Wordnet
• ConceptNet
• Sumo/Milo ontology

9
Data Sources II

• Dictionaries data retrieval
• On-line sources
• Merriam Webster
• Wordnet (gloss)
• MSN Encarta
• Approach
• Word morphing
• Phrases extraction (with POS tagger)
• Statistical analysis

10
Data access

• Binary dictionary search
• 220 1048576
• Binary search not acceptable in complex
  semantical applications
• Narrowing concept space by subsequent queries

11
20 Questions Game

• Algorithm

p(keywordvi) is fraction of concepts for which
the keyword has value vi Candidate concepts
O(A) are selected according to O(A) i
CDVi-A is minimal where CDVi is a vector
for concept i and A is a partial vector of
retrieved answers
12
Word puzzles

• 20Q game reversed
• Concept known
• Keywords the ones that would lead to the concept