Hinting Improbable WordLinkClouds for Cluster Summary and Discovery

1
Hinting Improbable Word-Link-Clouds for Cluster
Summary and Discovery

• Loui CS513 Proposal Fall 06

2
Problem

• There is too much to read
• We need skimming tools
• Text summarization is one way
• Improbable word and phrase extraction is another
• Word-Link-Clouds are also neat, e.g. Icosystems
  text visualization

3
(No Transcript)
4
(No Transcript)
5
Building These Things

• Which words to extract?
• Which words get links?
• How big each word?
• How thick each link?
• How can you HINT words?
• How can you VARY views?

6
Hinting

• Keywords are a great hint
• Ontologies such as wordnet are helpful
• Synonyms, meronyms, hypernyms
• Domain-specific dictionaries are helpful
• IBK the interest-background corpus

7

• Nouns
• hypernyms Y is a hypernym of X if every X is a
  (kind of) Y
• hyponyms Y is a hyponym of X if every Y is a
  (kind of) X
• coordinate terms Y is a coordinate term of X if
  X and Y share a hypernym
• holonym Y is a holonym of X if X is a part of Y
• meronym Y is a meronym of X if Y is a part of X
• Verbs
• hypernym the verb Y is a hypernym of the verb X
  if the activity X is a (kind of) Y (travel to
  movement)
• troponym the verb Y is a troponym of the verb X
  if the activity Y is doing X in some manner (lisp
  to talk)
• entailment the verb Y is entailed by X if by
  doing X you must be doing Y (snoring by sleeping)
  
• coordinate terms those verbs sharing a common
  hypernym
• Adjectives
• related nouns
• participle of verb
• Adverbs
• root adjectives

8
IBK

• IBK is a set of documents
• That are interesting
• Presumably contain keywords or their relatives
• Lack of representation Lack of interest
• Contains implicit information about granularity
• Lots of fine sports keywords IMPLIES detailed
  interest
• Variation/Repetition in IBK is as important as
  inclusion

9
Varying

• Icosystems/MIT/Bonabeau GA/GP feedback
• Levine/WUSTL my view, your view
• Loui/WUSTL Reordering a cluster creates a new
  view
• 513/WUSTL any new ideas?

10
Proposal

• Working just with BBC news data
• Write
• Visualization software (fairly trivial)
• Ordered clustering engine (many directions
  possible)
• IBK extraction/matching engine
• Multiple views

11
Needed

• People I can supervise who will do my bidding
• Scripting programmers OR very good UNIX people
• People who I can communicate with who can get
  things done quickly (not necessarily thoroughly)