For each item in a thesaurus, there is a list of similar things.

• Many of the similar things listed are so in peculiar or unexpected ways.
• Thesauri are usually built by finding written instances of usage of terms in particular ways.

Use a wiki to develop a design thesaurus.

• One page per entry.
  • A design term is usually more complex than just a word in a conventional thesaurus.
    • So, there should be at least a brief explanation of the term.
• Other terms that are analogically related to the one are listed, per conventional thesauri.
  • But, the other terms are wiki-linked to their own pages.

In conventional wikis, the links are defined by the context of the anchor.

• In page A, a link L to page B denotes a relationship between pages A and B.
• But the text directly associated with the link is usually just the name of page B.
  • It is not the relationship between A & B.
  • Notations like [Page name | visible text] do not explicate the relationship.
    • They are usually for grammatical consistency and composition.
    • They may help explain the relationship, but
      • this is not always the case, so
      • we cannot generalize the use of this notation usefully in this situation.
• So links in conventional wikis are basically untyped.

This means the wiki itself is unable to automate reasoning about the analogies (the relationships) themselves,

• because the analogies are only implied.

So what is missing is an ability to tag or annotate the link in page A to page B.

• If properly structured, this can provide the information needed by an automated system to work on the analogies themselves.
• This suggests that links would now be typed.

Example: a kind of searching that follows types of links between terms. It could produce chains of terms from some root term that result in lists of related terms.

• Of interest, the first and last terms in the chain could be extremely different.
  • This may suggest innovation design solutions.

How can such type information be added?

• This is suggestive of a semantic wiki.
• I can see two structural approaches:
  1. Build an ontology of link types a-priori, and
  2. Use a /tagging/ system to let the link types evolve through use.
• I prefer option 2.

Use WordNet plus tools like WordNet::Similarity to measure the distance between terms.

• Analogical reasoning

<refnotes> notes-separator: none </refnotes>