Search Done Right

One of the key aspects to presenting accurate and relevant search results is recognizing the user’s intent. A search for “IRA” on a financial services website may result in personal individual retirement account details as well as information about how to open an IRA or roll money over from a 401(k). A search for “cell” on Google may be a query related to cellular phones or it may be biology related.

These ideas have been getting a lot of buzz around the industry over the past year. Gartner analysts have labeled the need to improve refinement as “conversational search” – even predicting that by the end of 2010 more than half of new search projects will employ it. But the big question throughout the buzz is what does it really mean for the end user?

There are many ways to help an end user narrow down a search after the search has been executed, including spotlighting specific results, automatically creating categories or clusters of information based on the search results and allowing navigation of metadata information associated with the search results. 

The technique that allows the search itself to be modified or expanded before search results are displayed to the end user is called query expansion. Query expansion involves evaluating a user’s input (what words were typed into the search box and sometimes additional metadata) and then expanding the search query to include synonyms, spelling corrections, morphological forms of words (e.g. stemming) – including broader, narrower or related terms – and changing weighting of terms.

For example, a user may search for “paper,” which may be expanded to include synonyms and related terms. So the actual query may become “paper OR newspaper OR journal OR wallpaper.”

Many current approaches automatically expand a user’s query based on proprietary algorithms and a knowledge base that aren’t able to be changed by the end user. A search for “dog” may automatically be expanded to include “cat,” potentially leaving the user confused and frustrated over being presented with results that include only cat. The key features related to expanding the user’s query should include:

• Complete end-user visibility into what has been queried
• The user’s ability to control the query and choose the expansion terms
• The ability to interpret and expand both single words and phrases (e.g. “bodies of water” could be expanded to “oceans” and/or “lakes”)
• An administrator’s ability to change or add industry-, company- or application-specific query terms for query expansion
• Customization based on user profile information or application-specific information (e.g. where within the application the search is issued from)

Because it is not always possible to determine a user’s intent, a search application needs to be flexible and customizable by both the administrator and the end user as well as expose to the end user what type of query was executed. That’s one of the many ways to do search right.

This entry was posted on Monday, July 27th, 2009 at 9:50 am and is filed under Uncategorized. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.