Better Search with Predictive Typing: Can We Do it Better Than Google ?

Many of my favorite sites including google, linkedin, and netflix now have excellent predictive typing support. It strikes me that we should be able to do the same thing on SunSpace, and maybe do it better.In the case of Google, the suggestions returned will be what others are searching for. This list is  influenced  by your own search history.

The Netflix engine searches the universe of Movies and Actors. The list displayed is influenced by the popularity.

Your Linkedin type ahead universe is populated mostly by your connections.

We should be able to take the best attributes of the above!

On SunSpace we have more information to work with. Like Netflix we are searching a bounded universe. There are tens of thousands of movies and actors to select from, verses hundreds of millions of documents in the google index. In SunSpace we have less that 150,000 objects to work with, and lots of information on each object.

On SunSpace we know. (This is all tracked via opaque handles)

  1. Search terms used, for you and for all users on the system.
  2. The documents and wiki pages that were found via search and the search terms used, for you and for all users on the system.

Thanks to Community Equity we also know a lot about each user and each wiki page and document in the system.

  1. Information Value of each document or wiki page.
  2. Meta data for each document and wiki page. (author, tags, last modified, etc.)
  3. Equity value for each user. (based mostly on Information Value of the documents they own.)

As an example, let’s say we are searching for Cloud Computing. Google gives the the following list of recommendations:

Cloud Computing
Cloud Computing Companies
Cloud Computing Wikis
Cloud Computing Leaders
Cloud Computing Stocks
Cloud Computing Architecture

In SunSpace we can provide you not only with the recommendation of the search term, but also

  1. the most popular cloud computing documents people have found with search
  2. cloud computing documents with the highest information value
  3. interesting meta data for each document.

I believe that people generally utilize type ahead as a time saver, verses as a recommendation engine, so we need to make sure the “expected” results appear at the top of the list. If a user as typing in “cloud computing” in a prior search, that term, and the document the user selected from the results list should appear first.

As a bonus, we can match against the corporate directory in real time, and provide phone numbers and locations for individuals. This is normally a 10-20 second endeavor using the IT supported tool.

Posted at 06:09AM Dec 29, 2009 by Michael Briggs in Sun  |  Comments[0]