Who, What, Where & When: A New Approach to Mobile Search, IUI '08

http://delivery.acm.org/10.1145/1380000/1378817/p309-church.pdf?key1=1378817&key2=3643821721&coll=ACM&dl=ACM&CFID=81639924&CFTOKEN=12013848

Comments:

Summary:
The paper talks about a prototype for a new form of a mobile search engine.  It argues that mobile device users, due to typing difficulties, will always provide short and vague search terms on their mobile devices.  This means that the old searching model will become less capable of returning desired results.  To rectify this issue, the researchers explain that more context sensitive data should be included with the search.  Currently Google and Yahoo's mobile search takes into account the users location however, much more can be done.

The researchers created their own prototype for a search engine which takes into account more context sensitive data.  Their search engine utilizes past queries to help make generalizations about the users preferences.  The results page is a dynamic display based on this information.

As the figure shows the results are very dependent on physical location and previous searches.  The queries are shown by yellow markers where the results are shown by red markers.  The two sliders at the bottom will also affect the results of the qurery.  The top slider is based on time going from earlier to now.  This will base results on when queires were made.  The second slider goes from broad to specific, modifying how much the results are affected by past queries and the search engines guesses as to the users preferences.

Discussion:

I think something like this could either be really cool or really annoying.  The biggest problem with it is that it will guess wrong so many times people will always only use the now and specific slider positions, essentially rendering the added features useless.

I have noticed a little bit of difficulty in getting good results from broad terms on my phone and do see a need for some improvement.  The locational features used by Google is really nice and I have seen a lot of use from that.  It only seems natural to add more context sensitive data to the results we currently recieve.  Hopefully this will be translated into something great and usefull.