• by Kaila Colbin , Featured Contributor, November 2, 2007

Microsoft is the latest to take a stab at deciphering intent, outlining updates to Live Search in a pair of well-titled blog posts: "Do What I Mean, Not What I Say!" (part I and part II). The posts describe new semantic features that Microsoft hopes will take it further along the road to disambiguating user intent:

1. AutoSpell Correction Microsoft's description of this is kind of funny, because it makes a direct comparison to Google's treatment of suspected spelling errors without, of course, referring directly to Google:

   If we are absolutely, completely, totally, "no doubt about it" confident you misspelled one of your search terms, we automatically deliver a page that includes spell-corrected results, rather than a page of misspelled results accompanied by a "Did you mean _______?" link at the top... ...With AutoSpell correction I get the correct result the first time, regardless of the misspelling. Instead of being two clicks away from pizza, I'm just one. Being two clicks away just keeps people hungry, rather than satisfying their intent!

   Google, of course, has been checking spelling for years, but evidently the engine's agonizingly long two-click process of asking, "Did you mean..." has deprived searchers of much-needed pizza.
2. Stemming Stemming is about knowing when you mean "books" instead of "book" but not "cables" instead of "cable."
3. Equivalencies Equivalencies are abbreviations and other instances of words whose meanings were heretofore known to searchers but not to engines. The example Microsoft gives is "CA CHP," meaning California, California Highway Patrol (although why you'd need the redundant California is beyond me).
4. Intelligent 'Stop Word' Retention Stop words, like "a," "the," "in." etc., are typically discarded by Google, even though sometimes they add value. The example Microsoft gives is "The Office," which is an entirely different query than "office."

These features are aimed at what I call "structural semantics": the ability to ascribe meaning based on the structure of the query.

Laurie Petersen at Online Media Daily recently discussed a different approach from Acxiom:

Database marketing giant Acxiom today officially launches its Relevance-X products designed to allow marketers to make online media buys... "We're really excited about this," said Rich Howe, Acxiom's chief marketing and strategy officer. "We're bringing our knowledge and experience in direct marketing to the online channels to give clicks context -- going far beyond basic information such as age, gender and household income to include the attitudes, beliefs and lifestyles of consumers that are much more predictive." [emphasis mine]

According to the article, Acxiom has been seeing click-through rates double or triple in tests of the Relevance-X system: powerful numbers. The ability to successfully map a user's intention has tremendous implications for how people access the Internet, and it's a worthwhile exercise to imagine an Internet that responded and adapted directly to the individual, in real time. Imagine, for example, how your usage of StumbleUpon might change if you knew that StumbleUpon would deliver the exact Web page most relevant to you personally, every time.

The company that I blog for, VortexDNA, is also in the business of intention mapping, with technology reportedly based on the knowledge that an individual's purpose and values can be used to predict relevance.

Microsoft, Acxiom, and VortexDNA: we're all saying the same thing. Hugh MacLeod of gapingvoid says it too:

...it's not the wine per se that is interesting, it's the conversations that happen around the wine that is interesting. And that is true for all social objects. People matter. Objects don't.

You said it, Hugh. People matter. Not objects, or keywords, or stemming, or equivalencies. These are all tools to help us understand the people behind the queries and deliver what they want. When it comes to search, the road to heaven is paved with user intention.