The Mymory Project
Menu

LearningRulesForTheAcquisitionAndMaintenanceOfPersonalInformationModels

Diploma Thesis Frank Osterfeld

Overview

Starting point of this thesis is the personal information management of a knowledge worker on his workplace. In his daily work, the knowledge worker builds a mental model, a Personal Information Model (PIM) of his environment. This model comprises various concepts, e.g. the people he works with, the projects he is involved in, the specific tasks he is responsible for, documents and messages in electronic and non-electronic form, meetings he attended, and the relations between these concepts (e.g. co-workers work on a project, some people are attendees of a meeting). If the system had a model of the user's PIM, it could support his work in a better way, by providing e.g. appriopriate visualizations of the information on the system.

To obtain the user's PIM, we need to bridge the gap between the informal, implicit model in the user's head and the explicit, formal representation that is required by the system. The easiest approach, letting the user manually build and maintain a formal representation of his mental model, is too complex and tedious to be desirable. Thus, building and maintenance of such models should be automated as far as possible.

This thesis will explore the possible degree of automation in PIM maintenance. The focus will be on the maintenance of an existing PIM instead of building it from scratch.

The approach followed is to maintain the user's PIM by observing the user's interaction with native structures on his computer like addressbook entries, e-mail, bookmark and file folders. We assume that these structures provide a view of the user's PIM and actions on these structures indicate changes in the model (like participation in a new project when creating a new subfolder in „Projects“ folder).

As part of the thesis, a prototype will be implemented suitable to test the (semi-)automatic PIM maintenance in real-world scenarios.

The following steps are planned:

  • identify common patterns in the semantics of native structures and derive heuristics.
  • observe user actions performed on native structures using the EPOS/Mymory-Framework and use them as input.
  • use Machine Learning to adapt to the user and to suggest appriopriate model changes. The heuristics mentioned above can be used as prior knowledge.
  • semi-automatic approach: as the understanding of the user's PIM won't be perfect, let the user check and if necessary correct the suggested model changes: Present the suggestions in a „user-friendly“, human-readable way. Let the user easily confirm, correct or reject statements.
Go to top   More info...  

This page last changed on 01-Sep-2006 14:11:08 CEST by elst.