Gurungo: Coupling Personal Computers and Mobile Devices Through Mobile Data Types
Ivan GonzalezMicrosoft
Jason HongCarnegie Mellon University
1 – Theoretically, Incredible Access
1 – In Practice, Harder to Get Info
• Smaller screens• Slower text input• Slower network speeds
2 – Why Do We Print Maps?
2 – Why So Difficult to Get to Mobile?
• Synchronization tools useful for email, calendar, but still lots of useful information just thru browsing
3 – Re-finding Information is Common
• Tauscher and Greenberg 1997 found 58% of web activity was re-visiting old web pages
• Cockburn and McKenzie 2001 found that 81% of web pages were previously seen
• Obendorf et al 2007 found:– 72% of revisits happen within an hour
– 12% of revisits happen within a day
4 – Not All Information is Equal
• Sohn et al’s CHI 2008 diary study on Mobile Information Needs
• Lots of kinds of information useful when mobile– Trivia
– Directions
– Points of Interest
– Movie times
– Phone numbers
– Flight info
• In many cases, these kinds of mobile data typescan be automatically detected
Gurungo
• Make it easy to acquire and share data you already interact with on PC with mobile device
• Automatic Sharing– Implicitly monitor stream of web pages on PC
– Detect mobile data types
– Annotate the data (e.g. synthesized voice directions)
– Copy data to mobile device
• Manual Sharing– Copy and paste metaphor
Related Work
• Komninos and Dunlop 2007, pre-cache content based on calendar entries– Ex. name of atypical place in calendar, get maps
• Harding et al 2009, plan ahead and show manually entered information based on contextual triggers– Ex. Show travel info based on time
• With Gurungo, cache data that people directly interact with on PC, based on mobile data types– Variant of old idea of locality
Gurungo Overview
• Automatic– Implicitly monitor the stream of web pages
– Detect mobile data types
– Annotate the data (e.g. synthesized voice directions)
– Copy data to mobile device
• Manual– Copy and paste metaphor
• Two data types implemented– Driving directions
– Product details for price comparisons
Detecting Mobile Data Types
• Use a FireFox addon to monitor web pages– Goes thru the HTML DOM
– For predefined web pages, use XPath to get data
– For unknown web pages, use regular expressions and keywords
• We used a hybrid approach– XPath good for hard to specify data (maps, movie times)
– Regex and keywords good for broad coverage
Annotate Data
• Use web services or local programs to improve usability and/or utility of the data
• Driving Directions– Generate synthesized speech
• Product details– Get product reviews
– Get prices on web sites
Copy Data from PC to Mobile
• Proactively copy the data over to the mobile device– Currently, just keeps all info, no garbage collection
Manual Copy and Paste
Mobile User Interface – Directions
Mobile User Interface – Products
Discussion
• Static versus dynamic mobile data types– Driving directions and product details good for months
– Flight information good for … minutes?
– Traffic reports, social events, movie times, store locations
• User interface– Needs to be able to scale up more
– Possible to use location and recency to filter
• Garbage collection– Some data has natural expiration (social events)
– Other data does not, may opt to collect oldest and unused
• Lots of assumptions, need to verify with user studies– Re-finding info on mobile, recency of info
Summary
• Gurungo, a system for coupling PCs and mobiles based on data that people see and use on desktops– Not all data equally useful when mobile, bias UI
– Detects mobile data types based on what people already do
– Annotates data for usability/utility
– Make it easily available on mobile
Gurungo: Coupling Personal Computers and Mobile Devices Through Mobile Data Types
Ivan GonzalezMicrosoft
Jason HongCarnegie Mellon University