Why People Search for Images using Web Search Engines

摘要

What are the intents or goals behind human interactions with image searchengines? Knowing why people search for images is of major concern to Web imagesearch engines because user satisfaction may vary as intent varies. Previousanalyses of image search behavior have mostly been query-based, focusing onwhat images people search for, rather than intent-based, that is, why peoplesearch for images. To date, there is no thorough investigation of how differentimage search intents affect users' search behavior. In this paper, we address the following questions: (1)Why do people searchfor images in text-based Web image search systems? (2)How does image searchbehavior change with user intent? (3)Can we predict user intent effectivelyfrom interactions during the early stages of a search session? To this end, weconduct both a lab-based user study and a commercial search log analysis. We show that user intents in image search can be grouped into three classes:Explore/Learn, Entertain, and Locate/Acquire. Our lab-based user study revealsdifferent user behavior patterns under these three intents, such as first clicktime, query reformulation, dwell time and mouse movement on the result page.Based on user interaction features during the early stages of an image searchsession, that is, before mouse scroll, we develop an intent classifier that isable to achieve promising results for classifying intents into our three intentclasses. Given that all features can be obtained online and unobtrusively, thepredicted intents can provide guidance for choosing ranking methods immediatelyafter scrolling.