User Driven Search

From Project VRM
Revision as of 00:20, 17 May 2009 by Judi (talk | contribs) (added Asa Hardcastle's notes)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

User Driven Search (UDS)

Joe presents the concept of User Driven Search using a post that he made in July 2008.

principals of UDS see: http://blog.joeandrieu.com/2008/07/12/towards-user-driven-search/

Q. How Structured is a search map? A. UDS not bounded by structure but richer than keyword only. Excel as a metaphor: powerful without structuring the input.

RDFA: Can be used to describe any ontology. Google has recently adopted a specific ontology.

Bo from Infoseek: Holy Grail of search is a one or two word search. Infoseek would watch what people would click on after a search. Tweeked searches partially based on clicks after search.

Current problem: switch from searching for "leather" after searching for "Hawaii" - search engine tries to connect the two. Compartmentalizing is one of the values of a search map.

Problem with the current mindset is that the keyword is your search. We know that your keywords evolve as you refine your search.

File format for UDS is a search map. Map created by plugin, living in the browser context. As you search, a history is being visualized. Also keeps all of the pageviews. You can also delete items from your history that are not relevant. It is also possible to grab content and add it to your map.

Motivation for collecting portions of your search into your search map is because as you search, you find good stuff that you want to keep.

SwitchBook algorithm: takes what is in the search map and generates recommendations.

  • organic results based on other people's search maps that are similar - bubbles up things that you have not seen, can delete these at any time.
  • there is a log of all of the recommendations so that if you missed something you can go back and look.
  • you can set up filters on recommendations (no ads, no sony, no...)

Entire scope of identity is in a specific search map.

High Fidelity representation of your search intent. Can't give social representations unless a lot of people are using the SwitchBook technology.

OSS:

  • Server Reference implementation
  • Client browser plugin

API:

  • there will be a lot of apis

Starting in IE will get to firefox as quickly as possible.

You will be able to customize the client views using CSS/Javascript Tiddlywiki code. There will be a way for you to load in your own tiddlers.