Selecting versus Describing: A Preliminary Analysis of the Efficacy of Categories in Exploring the Web
E. G. Torns, Rick Kopak, Joan C. Bartlett, Luanne Freund · Text REtrieval Conference · 2001
Summary of Results The 48 participants spent about 7 minutes doing each task. They used the search box forabout 66% of the tasks and selected from the directory categories for the remainder. On average,they examined about 5 URLs and about 6 links within each of those URLs. They tended to selectabout the fourth item on a hitlist and on average examined about two pages of hitlists. Participants reported little familiarity with the topics for each of the assigned tasks, with fewhaving ever done a search on any of the topics prior to the session. On a five-point scale with onebeing the poorest rating and five being the best rating, they indicated the degree of certainty withwhich they found their answer, the ease of finding the answer, and their satisfaction with theprocess of finding their answer at around four. User-Specified vs. Researcher Specified Task Half the questions were completely specified and half were fill-in-the-blanks, allowing someuser modification toward personalizing the task. There were no significant differences between thetwo types on any measure. This finding challenges the assumption that information retrievalexperimentation with pre-defined queries alters user behaviour in experimental settings. Ourparticipants performed about the same regardless of whether they were assigned a task orallowed to create their own. That said, it is likely that the artificially of the process, e.g., timeconstraints, lab setting, and so on, may have a greater impact than the nature of the task.