“Method and system for autocompletion using ranked results” is a patent granted to Google on the 3rd of February 2009.
Their invention is to process query information received from a user prior to the user having indicated that the query is completed. So just as we have seen with the suggested searches appearing below the query box, this allows for your query to be autocompleted based on a number of variables.
It predicts the queries ordered by a ranking algorithm. These are then served up to the user. The complete queries that were submitted by a community of users who also searched for things corresponding to the partial query, so history is a factor. ”…each of the complete queries in the set of historical complete queries having frequency of submission information associated with the complete query;” A number of
The user enters further information if the predicted query is not correct and continuing to do so allows for a complete query to be enetered, or until a new set of predicted queries are shown.
You would enter a succession of search terms until you are satisfied that all of them are entered and then signal to the system that you have finished. The engine processes the query and a list of results appears.
The do mention that one drawback is that because the engine has to wait until the user has enetered all the search terms before the system can processit, the method is slower than is desired.
This is an interesting development because query suggestion becomes more personlaised, seeing as the community of users with the same query type and who recieved results relevant to these would affect your own results. This type of system can be observed in other research as well, which is due to be presented in the next couple of weeks. Using information from other social resources could be an interesting development.