a drop of water

when enough gathers, you have to fall somewhere

Associative Word Geometry Applied to Internet Searches

View Comments

Every word has an association with every other word.  The associations can be strong or weak and negative or positive.  i.e. Love has a strong positive association with romance and a strong negative association with hate.  One could theoretically assign a point in an n-dimensional space (where each dimension is a type of association).  By plotting each word in a search string in this space and finding planar intersections, a search engine could return a set of results of higher quality based on the users search.  The advantage of this algorithm would be that it factors in the associative relationships of words by all users in order to interpret the intent of a single user.

Written by ddhoffman

December 1st, 2008 at 12:09 pm

Posted in Ideas,Tech

Tagged with , ,

«
»
  • Well, I suppose the computational complexity of the algorithm requires a complex search to make full use of it so its hard to come up with a viable example with just two words. The complexity may not even be worth the benefit. I suppose the algorithm could be given a learning aspect, where results of like minded people could be grouped together, so that search results wouldn't have to be limited in accuracy to the general population's word associations.
  • Wow, for a post on love and romance, this sure got a little weird.

    Lots of big words - I need an example to fully understand it. Well, ok, I understand it, but what if I associate chap stick (first thing I saw, examples are a little weak today) with lips, and someone else thinks medicine? What will Danoogle provide as the number one result? If people think unlike I do (which is a given), doesn't this have just as great a possiblity to muddy the search waters as purify them?

    PS. Just call me the Dream Crusher
blog comments powered by Disqus