Posts tagged ‘or’

AND or OR

I’ve just re-found this note that a former colleague wrote. A useful reference point.

The basic premise is if you do an ‘OR’ then you get the ‘and’ results at the top and the ‘OR’ results follow them with lower relevancy as they only have one of the words. This then allows you to maybe get a decent result returned even if there are no results with both words in the search.

Consider a search for “money euro emu bank of England”, with an ‘and’ search you would not get back results which contain 4 of the 5 words in the query term, and there may well not be any docs with all 5 words in. So if this were an ‘and’ search by default then basic search would not give you back any docs even though some may be highly relevant to the person searching. The person searching in this example has tried to be as helpful as possible by giving lots of words that are relevant to the kind of doc he is looking for probably without intending to restrict his search.

If you look at it from the other point of view then why would you do an ‘and’ search as the basic search instead of an ‘or’ search? The only argument I have heard is that it doesn’t clutter results with ‘or’ docs and doesn’t give back as many docs so is less likely to confuse the user. I don’t understand this because if you trust the search engine and the algorithms behind it then the relevancy will sort everything so that all the ‘and’ docs that have all words in the doc returned will be at the top of the results when doing an ‘or’ search anyway because all words hit so gets a higher score than if only one or two words hit. If you do a search using Google you will be returned umpteen million docs but its not confusing because the relevancy algorithm sorts them so that the most relevant docs are in the top 10 based on all words in the query term hitting and hitting most frequently.

12 September 2008 at 23:08 Leave a comment


Categories

August 2017
M T W T F S S
« Sep    
 123456
78910111213
14151617181920
21222324252627
28293031  

Twitter Updates