June 17, 2007

Search Results that Enhance a Text

Google has an interesting patent that describes the philosophy behind Google AJAX Search API (also known as user-distributed search), a way to bring the search results to the user.
In an increasingly networked world, users frequently use online sources to create and exchange information. Email, instant messaging (IM), message boards, websites, and blogs are all existing communication technologies through which users can create and distribute content to other users. Frequently, in creating such content, a user may wish to reference other online information sources. For example, a user authoring an email may use a browser to navigate to a web page that the user would like to reference in the email, copy the link (e.g., the uniform resource locator (URL)) from the browser to a "clipboard," and then paste the link from the clipboard into the email. In this manner, the user can create an email message that contains links that are accessible by an eventual reader of the email.

Search engines are a popular tool through which users enter a search query describing information of interest and receive back documents or links to documents that relate to the search query. Frequently, when "researching" content for an email message, IM message, message board post, website post, or blog post, the user may perform one or more searches using one or more search engines to locate online documents relevant to the content. The user may then copy a link into the document using the above-described method of copying and pasting a link to the document. This process for annotating user created content can be tedious, difficult to perform for average users, and often results in textual links in the final content that can be difficult to read.

The patent suggests implementing a sidebar that lets you perform searches and easily include the results in the post or message. Another interesting idea is to show implicit search results based on what you type.
Instead of waiting for a user to provide a search query, the UDS may automatically generate search queries based on, for example, entity recognition techniques performed using the content entered by the user. (..) Entity recognition techniques are generally known in the art, and may include, for example, techniques designed to recognize entities such as products, places, organizations, or any other entities that tend to be subjects of searches. The entity recognition techniques can be based on linguistic grammar models or statistical models. In one possible implementation, the entity recognition techniques may be particularly adopted to locate terms that correspond to commercial products or terms that define an address, such as a postal address. In other possible implementations, the entity recognition techniques may be particularly biased to locate terms that are associated with a profile of the user, such as profile explicitly generated by the user (e.g., by the user filling out a questionnaire) or a profile automatically generated for the user, such as a profile based on the user's search history or based on documents created by the user.

This could be the first step towards a rich text editor smart enough to suggest relevant information for what you write and to auto-complete recurrent titles, names or ideas. One of the best existent applications for user-distributed search is Linkify, a bookmarklet that lets you place links in a text box using navigational queries.

{via SEO by the Sea}

3 comments:

  1. I use Linkify (per your suggestion in an earlier post). It's great! Now for example if I want to cite something about Satan , you'll know exactly who I'm talking about.

    ReplyDelete
  2. But will it blend?

    haha just trying Linkify... it's awesome! i wish it worked in gmail though!!

    ReplyDelete
  3. At first reading it sounds terrible. Like the bloody paper clip in 'Office'.

    ReplyDelete

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