Abstract

We describe an initial study of how participants perceive information when they categorize highlighted textual units within a document marked for a given information need. Our investigation explores how users look at different parts of the document and classify textual units within retrieved documents on 4-levels of relevance and importance. We compare how users classify different textual units within a document, and report mean and variance for different users across different topics. Further, we analyze and categorise the reasons provided by users while rating textual units within retrieved documents. This research shows some interesting observations regarding why some parts of the document are regarded as more relevant than others (e.g. it provides contextual information, contains background information) and which kind of information seems to be effective for satisfying the end users (e.g showing examples, providing facts) in a search task. This work is a part of our ongoing investigation into generation of effective surrogates and document summaries based on search topics and user interactions with information.