Some IdeasQuestions I am Thinking about

1
Some Ideas/Questions I am Thinking about

• Chiao, Hsin-Chen
• 07/26/05

2
Outline

• Introduction to Information Gain
• Classification
• Polysemy and Synonym
• The Taxonomy Generation

3
Classification

• Motivation
• Our binary classifier LiveClassifier
• Solution
• A new classifier that can classify a NE into
  Nation
• Approaches
• Concept-based and instance-based

4
Concept-Based and Instance-Based Taxonomy
CS
Nation
AI
Graph
Theory
Asia
Africa
Europe
5
Concept-Based and Instance-Based Taxonomy
Car
CS
GM
Ford
Jaguar
Car
AI
Graph
Theory
Racing
Truck
Jeep
6
Co-occurred Terms

• Which one gets the higher probability that the
  class names co-occur with queries?
• Who builds the taxonomy tree? Can it be automated?

7
Fundamental Questions on Classification

• What?
• Why?
• Dewey Decimal Classification System
• News classification
• Observation Classification based on some common
  property and classified with other different
  properties
• Is the Nation Classifier a Classifier or a
  Labeler?

8
Polysemy and synonym

• Query expansion
• Vector Space Model term independent
• Word Sense Disambiguation (WSD) discusses about
  choosing the most common sense of a word
• How to solve the polysemy problem?
• Expand vector length?
• Using basic dictionary words?

9
Questions about Polysemy

• How do we know which meaning the word is
  representing now?
• Can we interpret a word by its neighbor words?

10
Questions on TG

• Can we build a taxonomy tree according to the
  basic dictionary words and classify terms into it
  (term-based)?

MSN, Google, Map, Search
a, b, c
a, b
Summarization, Keyword Ext (IG), Dictionary words
c
d
MSN, Image, Google, Apple
a, b, d