March 5, 2007

How Google Desktop Ranks Search Results

A patent application recently filed by Google ("Temporal ranking scheme for desktop searching") gives some idea about how Google Desktop indexes and ranks files. The main criteria for ranking files is how likely a file will be accessed in the future. To estimate this value, Google Desktop tries to find different hints about the importance of a file.
It would be desirable for a user to be able to access applications, desktop files, and/or data on local or networked servers through a single interface. Access to each of these could be initiated by finding and identifying the correct resource, through the user's selection of a search result. Search results should be generated using methods that reflect how desktop resources are used. In most cases, the most relevant items are likely to be those that have been accessed or used recently and frequently, or have been designated as particularly important by the user, for instance through the creation of a desktop shortcut. (...)

To make the results of the searching more useful to a user, the items being searched are ranked according to their relevance, and based on this ranking, the items can be selected and/or ordered for display to a user. A temporal ranking scheme ranks the search results according to an algorithm designed to give higher scores to items that are more likely to be desired by the user in the search. In one embodiment, the relevance of a particular search result is determined according to an estimation of the frequency that the corresponding item will be used in the future.

In one embodiment, a group of items are ranked based at least in part on a last access time and a class associated with each item, where the class associated with an item indicates a frequency of use of the item. Possible classes include a location of the item (e.g., whether the item is a document in a "MyDocuments" folder), a type of the item, and a file size of the item. Once the items are ranked and a search is performed, the results from the search (e.g., the matching items) can be displayed according to their rankings, where the order of the displayed results may be based on the ranking and only the highest ranked results may be displayed. In this way, the search, such as a desktop search, returns the items matching the search query that are more likely to be the items for which a user is looking.

{ via SEO by the Sea }

5 comments:

  1. So, google has officially entered the frivolous patent game?

    ReplyDelete
  2. This seems counter-intuitive to me. If something is important and I have a desktop shortcut to it, why would I be searching for it? I only use desktop search for the really obscure things I only vaguely remember. Maybe they'll have a way to reverse the order so I can find what I couldn't find just by browsing the folders I'm always looking through anyway.

    ReplyDelete
  3. The point, or ultimate intention, is that you stop clicking on icons almost entirely.

    You have a shortcut for 'todo list.txt' on your desktop. Instead of going through /all that effort/ of remembering its on the desktop, locating it on the screen and moving the mouse to click on it, instead of all that you hit ctrl twice (or whatever the shortcut is) to pull up google search, type 'tod' and hit the autocomeplete button which will instantly open the file.

    Very useful and a completely different way of using the computer, or rather a return to CLI but a CLI with intelligence.

    I would however have thought this was entirely obvious. Isn't this a simple leap from Archy?

    ReplyDelete
  4. I forgot to mention that, by default, Google Desktop orders search results by last modified date. That means sorting results by relevance wasn't good enough to become the default option.

    ReplyDelete
  5. I am searching in a way that i can increase the relevance of outlook contact so that when i type any name the first result should be the contact name.

    ReplyDelete

Note: Only a member of this blog may post a comment.