Towards Fine-Grained Adaptation of Exploration/Exploitation in Information Retrieval

Abstract

Lookup and exploratory search tasks can be distinguished using individuals' information search behaviour. Previous work, however, has treated these search tasks as belonging to homogeneous categories, ignoring the specific information needs between users and even between search sessions for the same user. In this work, we avoid this dichotomy by considering each search task to exist on a spectrum between lookup and exploratory. In doing so, our approach aims to dynamically adapt exploration and exploitation in a manner commensurate with the user's individual requirements for each search session. We present a novel study design together with a regression model for predicting the optimal exploration rate based on simple metrics from the first iteration, such as clicks and reading time, that can be collected without special hardware. We perform model selection based on the data collected from a user study and show that predictions are consistent with user feedback.