Title: From documents to tasks: deriving user tasks from document usage patterns
Authors: Oliver Brdiczka
Venue: IUI '10 Proceedings of the 15th international conference on Intelligent user interfaces
Comments
http://detentionblockaa32.blogspot.com/2011/04/paper-reading-22-personalized-user.html
http://vincehci.blogspot.com/2011/04/paper-reading-22-pomdp-approach.html
http://vincehci.blogspot.com/2011/04/paper-reading-22-pomdp-approach.html
Summary
The focus of this paper was on increasing the efficiency of multitasking in the average workplace. Existing systems for task management require a large amount of time investment from users to be effective, because the systems need to be 'trained' in order to work correctly. This paper proposes a new approach for automatically estimating a user's tasks from document interactions, without requiring access to the content of those documents. The system described in the paper instead looks at only the switches between which document is being interacted with by monitoring each user's activities and logging which documents had focus and when. Comparing this data allows the authors to build a similarity matrix based on document focus frequencies, dwell times, and switches. A clustering algorithm is then used to group documents into tasks based on the similarity matrix.
To evaluate the performance of the system, three values were used, called prevision, recall and F-measure. Precision refers to the faction of documents in a cluster that belong to the task label of that cluster, recall represents the fraction of all document that belong to a task label and appear in the corresponding cluster, and F-measure is the weighted mean of precision and recall. These values are graphed to the left.
The focus of this paper was on increasing the efficiency of multitasking in the average workplace. Existing systems for task management require a large amount of time investment from users to be effective, because the systems need to be 'trained' in order to work correctly. This paper proposes a new approach for automatically estimating a user's tasks from document interactions, without requiring access to the content of those documents. The system described in the paper instead looks at only the switches between which document is being interacted with by monitoring each user's activities and logging which documents had focus and when. Comparing this data allows the authors to build a similarity matrix based on document focus frequencies, dwell times, and switches. A clustering algorithm is then used to group documents into tasks based on the similarity matrix.
Discussion
The system described in this paper seems promising. The authors described existing systems as either deficient, or requiring access to the content and title of each document in order to derive similarities from them, which raises security concerns in some businesses. It would be interesting to see how this approach compares to other systems that do analyze the content of each document in order to determine if the patterns associated with document access is sufficient to build similarities, or if more information is needed.
No comments:
Post a Comment