Despite to the vast advances in capacity and capability of computers over the past two decades, the representation of users, their tasks, activities and preferences in application software has changed little since the dawn of modern personal computing and personal information management (PIM) tools in the 1980's. This has left applications with a deficiency of appropriate information about users, which, when compared to the quantity and diversity of information managed for users by applications today, has created a fundamental bottleneck with respect to how well modern application software can serve usersù tasks and informational needs.

In this project, we propose a strategy for enriching end-user applications with information about their users obtained using three simple strategies: mining information already available on a userùs own personal devices, logging user activity and contexts unobtrusively through these same devices, and by opening up channels by which users can easily and flexibly express knowledge to the system as part of his or her workflow. Using this information, we demonstrate ways by which systems can provide cognitive, interaction, automation and task -oriented support to users that can save users time, effort, and frustration. Finally, we demonstrate how client-side user modeling approaches, unlike centralized collaborative approaches prevalent today, can scale to employ a vast quantity of private, sensitive information without sacrificing user privacy.

PLUM - a framework for client-side

PLUM, a software platform for Personal Lifetime User Modeling is our framework for exploring how long-term models of users situations and activities can be built and applied to personalizing interaction. PLUM is designed to interface directly with the operating systems and applications software running on the digital devices most people use today.

The following is an incomplete list of the components of PLUM:

plum.Chon

Daemon for activity and context capture, that watches the user's activity and situational context builds a structured chronology of its obserations using RDF. Chron relies on pluggable knowledge sources to enable interfacing with other platforms and new applications.

plum.UI

Components for controlling Chron and visualizing contents of Chron in real time

plum.model

State model abstraction for modeling managing representation of context and activity

plum.nlp

Code for context-based entity resolution and parse disambiguation

plum.peel

a javascript command shell that lets you interactively query the chronology, build classifiers, register listeners, or change the state model

Implementation features Papers

Van Kleek, M., Michael Bernstein, David Karger and mc schraefel. Getting to Know You Gradually: Personal Lifetime User Modeling (PLUM) Research abstract, MIT CSAIL, April 2007.

Van Kleek, M., Howard Shrobe. A Practical Activity Capture Framework for Personal Lifetime User Modeling (PLUM), student poster, 11th International Conference of User Modeling (UM 2007); Corfu, Greece; June 2007.

Bernstein,M., Max Van Kleek, David Karger and mc schraefel. Management of Personal Information Scraps. work-in-progress poster, CHI 2007, San Jose, CA, 2007.

Van Kleek, M., Howard Shrobe. A framework for ubiquitously capturing users' work and life practices. Extended abstract, CSAIL Student Workshop. Cambridge, MA, Sept. 2006.

Posters

Nokia-MIT Day poster: Achieving more personal personal information mangement. [ pdf ]

Releases

Update (1.18.2008) - PLUM is undergoing rapid active development but some pieces are still incomplete ; thererfore we are not yet ready for a general release. plum will be released under an MIT- or GPL license to encourage further development by others. Please contact the PLUM team if you are interested at plum at csail dot mit dot edu.

For more information

See the (yet incomplete) user's manual and contact plum [at] csail dot mit dot edu if you are interested!

Acknowledgements

PLUM is part of the ConnectingME/Virpi initiative at the Nokia Research Center Cambridge and the Web Science Research Institute.