[DigLib] Stanford Digital Library Project [Working Papers]

SIDL-WP-1997-0067

Exploring versus Exploiting when Learning User Models for Text Recommendation

Marko Balabanovic

marko@cs.stanford.edu

Abstract: The text recommendation task involves delivering sets of documents to users on the basis of user models, which are improved over time given feedback on the delivered documents. When selecting documents to recommend, a system faces an instance of the exploration/exploitation tradeoff: whether to deliver documents which are known to match the user model learned so far, or to deliver those about which there is little certainty. In this paper, a large-scale simulation is constructed to investigate the effects of this tradeoff on the rate of learning user models, and the resulting compositions of the sets of recommended documents, in particular World-Wide Web pages. Document selection strategies are developed which correspond to different points along the tradeoff. Using an exploitative strategy, our results show that simple preference functions can successfully be learned using a vector-space representation of a user model in conjunction with a gradient descent algorithm, but that increasingly complex preference functions lead to a slowing down of the learning process. Exploratory strategies are shown to increase the rate of user model acquisition at the expense of presenting users with suboptimal recommendations; in addition they adapt to user preference changes more rapidly than exploitative strategies. These simulated tests suggest an implementation for a simple control that is exposed to users, allowing them to vary a system's document selection behavior depending on individual circumstances.


Note: Papers in this series are in development and are not in a final form for publication or general dissemination. They are subject to change. Please do not quote or further distribute them without explicit permission from the authors.
This paper was created on: 07/10/97 and last revised on:9/30/1997

Author's Comments: Submitted to User Modeling and User Adapted Interaction

Status: PRIVATE

Click here to see the full text of SIDL-WP-1997-0067 (PS)

Revision History

VersionFormat DateComments
5PS9/30/1997Submitted to User Modeling and User Adapted Interaction
4PS9/18/1997Almost complete version to be submitted to User Modeling and User-Adapted Interaction, special issue on Machine Learning for User Modeling, comments welcome.
3PS8/19/1997This is a draft of the first section of a paper. Although it is self-contained its purpose is to set the scene for experiments which are under way but not yet completed.
2PS8/19/1997This is a draft of the first section of a paper. Although it is self-contained its purpose is to set the scene for experiments which are under way but not yet completed.
1PS7/16/1997This is a draft of the first section of a paper. Although it is self-contained its purpose is to set the scene for experiments which are under way but not yet completed.

[Stanford] [DigLib]