Books' Interest Grading and Fiction Readers' Search Actions During Query Reformulation Intervals

Abstract

We compared fiction readers' search actions during various query reformulation intervals. We aimed to understand how readers' search actions differed between successful and unsuccessful QRIs and which search actions predicted the selecting of very interesting novels compared to less interesting ones. We conducted a controlled user study with 80 participants searching for interesting novels. Three types of browsing tasks and two types of catalogs were used. Our results demonstrated that browsing task type was associated to readers' document viewing behavior in terms of observed search result pages, opened book pages and dwell time on book pages. When browsing for topical novels, most effort was required to select somewhat interesting novels. When browsing for good novels, most effort was required to select very interesting ones. Logistic regression analysis yielded that the most significant predictors of higher document value were the number of observed search result pages and opened book pages.