When you type a query into a search box at Google or Yahoo or Bing on your desktop computer, chances are a drop down listing of suggested query terms will appear below the search box.
If you use a smart phone, and start typing into a text box on your phone, your phone may also offer you some suggestions to complete the word you are typing.
In the case of a cell phone where you need to press numbers to represent alphabetical characters, those suggestions can help save you from typing a lot of keystrokes. The phone offers terms from a dictionary stored on your phone to help you complete those terms.
A recent patent application from Google describes how they might add words to a dictionary like that, taken from social networks where you might be a member. What’s interesting about that is how much information the search engine captures about your use of words on the Web, and that of people whom you might be connected to on the Web.
Why might Google look to social network information for this kind of information?
The patent tells us:
The theory is that a user is more likely to use terms that their friends often use. For example, if a teenager has identified various users as friends on a social networking web site, the content of those friends’ pages and other similar content may be analyzed in determining popularity of terms for the user.
Such a user, for example, may be much more likely to use certain forms of slang in their communication–something that would not be picked up by a dictionary that is premised on more general usage of terms across a wider population.
The patent application is:
Invented by David P. Conway and Andy Rubin
Assigned to Google
US Patent Application 20100114887
Published May 6, 2010
Filed October 17, 2008
The subject matter of this specification can be embodied in, among other things, a computer-implemented method that includes receiving a request to provide a dictionary for a computing device associated with a user; identifying word usage information for members of a social network for the user; and generating, with the word usage information for members of the social network, a dictionary for the user.
The text associated with a member of a social network might include content such as:
- Pages on which they post information,
- Profile pages from places such as Orkut or Facebook or MySpace
- Discussion pages or text message logs of communications between the members of a social network.
The patent filing provides details on how they might score different words used by members of a social network to decide which words to add to a dictionary. This score aims at predicting words that someone might use in the future.
As an example, we’re told in the patent filing that:
If the users are teenagers, the analysis may identify many phrases that would not have appeared in a review of standard English usage, such as OMG (“Oh my God!”), “like,” “totally,” “sick” and other such slang terms.
In addition to helping phone users complete words taken from a dictionary, the patent filing also describes how the information from someone’s social network could be used in a search at a search engine to offer suggestions of terms to use.
This method for deciding upon words to present as suggestions could also look at information found from someone’s computer, taken from word processing documents, calendar items, contacts, history from a browser, and more. So, if someone frequently visits the baseball pages at ESPN and those files are in their browser’s cache of temporary internet pages, when they start spelling b-a-s, the computer they are using might offer “baseball” as a query suggestion.
Information located on a network that someone uses, such as their email account, might also be a source of data that could be used in helping someone fill out a text box on their phone, or in suggesting a query term.
I’ve written a few posts before about patent filings from the search engines on predictive queries:
- Can Google Read Your Mind? Processing Predictive Queries
- Google Improving Mobile Search
- Google predicting queries
- Yahoo’s Predictive Queries, Invisible Tabs, and Temporal and Monetization Bias Experiments
- Predictive Queries versus Unique Searches
- Yahoo’s “Universal Search” and Vertical Search Suggestions
- Predictive Search Query Suggestions
At least one of the patent applications those posts describe hints at the possibility that words shown as predictive queries might be taken from a group that someone is a member of, but it didn’t provide much detail.
Interestingly, this patent seems to focus more upon the auto-completion of words on a phone than it does upon providing query suggestions to a searcher. Given Google’s entry into smart phone software, that shouldn’t be a surprise. The patent filing also includes software and hardware details about a phone that might use a dictionary like this, and the images from the document show details about a phone running Google’s Android system.
What I find most interesting is how much attention Google may be paying to the words we use in our conversations on the Web.