Do you want to browse your past? Can you remember the significant work and personal events in your own personal timeline?
Microsoft Lifebrowser catalogues everything from your past to remember landmark events in your life and work, zooming in and out of the details using a 'volume control'
Eric Horvitz, Distinguished scientist at Microsoft research has been looking at machine learning to better understand how people think, using inferences about people to enhance their lives.
Behind the scenes, there is a machine learned model that reasons about each dimension of information in your stream such as images, documents and videos. This works out which events throughout your past are likely to be landmark events.
It uses a Bayesian model which works out the probability of the event being a landmark event from the user's point of view.
Life landmarksLifebrowser learns to predict 'memory landmarks', the events of life that you would consider important. It uses these landmarks to help you navigate large personal stores of information.
It uses information such as photos, search and browsing activity and appointments and collates them into a timeline. You can navigate the timeline of events, and with a 'volume control', you can opt to see more or less of everything in your timeline and expand it into as much detail as you want to.
You can search on anything from your timeline, and bring up the appropriate content for the items. You can see emails, documents and other items related to the search term.
What is really clever is that you can also see other things that were occurring at the same time as the event you search for. You can see personal information, as Lifebrowser has picked up on the fact that there was a landmark event around the same time. It has connected the two pieces of information.
This is not the first aid to memory research that Horvitz has worked on. In 2004 the team were working on other memory products:
They developed an image crawler that analyzes photos in a user's personal photo library. The image crawler collects properties of the images from information stored by digital cameras, and also by automatically extracting features of pictures with image analysis algorithms. This data is used to build a Bayesian model of which pictures would likely serve as memory landmarks.
You can search for a landmark event and using the landmark, drill down into other things that were happening at that specific time. Work, leisure, home, personal events are all captured in Lifebrowser. Over time, it analyses and learns what is important to you to help you remember your past.
Wider applicationLifebrowser could be useful for adding context to key decisions that you have made in your career, but it could be applied in more ways than that. Small businesses could use it to track their success from one man band to the creation of their large organisation, fixing dates with news events and business occurrences.
It could also be really useful to help those who struggle to remember key things in their lives. Machine learning could help patients who are in the early stages of dementia to catalogue stages of their lives for the next generation to cherish.
Correctly applied, Lifebrowser could be much more useful than Facebook's timeline jumble. This could bring context and reminders to our past -- providing we have the hardware capacity, or space in the cloud to store the minutiae of our lives...
- Microsoft showcases new Kinect-centric projects at its TechFest
- Microsoft shows off NUI, Kinect-focused research projects
- Microsoft goes public with its 'Drawbridge' operating-system
- Crowdsourcing to predict the future of computing
- Microsoft prototypes speaking multiple languages with a single human voice