Predictive text input in a mobile shopping assistant: methods and interface design, IUI '09

http://portal.acm.org/citation.cfm?id=1502650.1502714&coll=ACM&dl=ACM&type=series&idx=SERIES823&part=series&WantType=Proceedings&title=IUI&CFID=76752576&CFTOKEN=55465958

Comment:
Patrick Webster

Authors:

Petteri Nurmi  Helsinki Institute for Information Technology HIIT, Helsinki, Finland Andreas Forsblom  Helsinki Institute for Information Technology HIIT, Helsinki, Finland Patrik Floréen  Helsinki Institute for Information Technology HIIT, Helsinki, Finland Peter Peltonen  Helsinki Institute for Information Technology HIIT, Helsinki, Finland Petri Saarikko  Helsinki Institute for Information Technology HIIT, Helsinki, Finland The authors created a predictive text system for generating shopping lists on a mobile phone.  The system attempted to leverage the fact that people generally buy certain items together such as macaroni and cheese and ketchup, or peanut butter and jelly.  The system would look at what is already on the list and use these items to produce better suggestions for the user.  These suggestions were presented when the user pressed one key to start the word so if peanut butter is already present on the list when the user presses J, jelly will be presented at the top of the list instead of something else such as juice. A user study was done with the system where half the participants were given the predictive text system and the other half were given a standard system with no predictive text support.  The authors found that there was a significant increase in the speed using their predictive system, roughly five words per minute faster.  Some users were asked to user two hands to provide input while others were instructed on only use one hand.  This was to simulate the users at home, with two hands, and then in the store, with one hand.  They saw differences in speed between one handed and two handed use however these were both faster than the standard input method. I would like to see this application ported over to the Android and iPhone systems.  They were implementing it on a Nokia "dumb phone," but it doesn't seem like it would be that difficult to port over.  I liked the concept.  Anything that helps me save time and do more is good by me.  Predictive text is alright as is but adding the extra contextual element seems like it would make a much more refined system.