Wednesday, September 12, 2007

Providing the Ultimate Search Experience - Part II



We all know how google.com somehow miraculously finds the best documents for us - but did you know that a big part of that is based on a large synonym dictionary that was automatically generated using advanced Machine Learning techniques? One of the great advantages of owning a Google Search Appliance is that you get the benefit of what we've learned from billions of queries about words and their synonyms right in your box!

Query Expansion is a feature of the Google Search Appliance that intelligently expands the terms of the query based on the context to find more candidate documents. Query Expansion uses word stemming techniques (e.g . expanding "car" to "cars") and a list of common synonyms from Google's internal dictionary (e.g. expanding "car" to "automobile") to find all relevant documents. This feature can also optionally expand the query terms based on a custom dictionary that you upload to the appliance.

This feature is off by default. In most cases, you will usually see a marked increase in relevancy simply by enabling this feature.

There are 4 modes of Query Expansion from which to choose:
  • None: Disables query expansion
  • Standard: Enables query expansion using Google's built-in synonyms
  • Local: Enables query expansion using only the synonyms that you upload to the appliance
  • Full: Enables query expansion using both Google's built-in synonyms and the files that you upload to the appliance
So, if you have not uploaded any of your own custom synonyms to the appliance, then you should choose Standard. However, if you have uploaded your own custom synonyms to the appliance (highly recommended for company-specific colloquial or jargon), then choose Full.

Don't be shy... turn it on and try it out! To test more deterministically the effects of query expansion, stay tuned for a blog post on possible search quality evaluation techniques.

Happy expanding!

No comments:

Post a Comment