Associating search and navigation behavior through log analysis

Abstract

We report on a study that was undertaken to better understand search and navigation behavior by exploiting the close association between the process underlying users' query submission and the navigational trails emanating from query clickthroughs. To our knowledge, there has been little research towards bridging the gap between these two important processes pertaining to users' online information searching activity. Based on log data obtained from a search and navigation documentation system called AutoDoc, we propose a model of user search sessions and provide analysis on users' link or clickthrough selection behavior, reformulation activities, and search strategy patterns. We also conducted a simple user study to gauge users' perceptions of their information seeking activity when interacting with the system. The results obtained show that analyzing both the query submissions and navigation starting from query clickthrough, reveals much more interesting patterns than analyzing these two processes independently. On average, AutoDoc users submitted only one query per search session and entered approximately two query terms. Specifically, our results show how AutoDoc users are more inclined to submit new queries or resubmit modified queries than to navigate by linkfollowing. We also show that users' behavior within this search system can be approximated by Zipf's Law distribution. IntroductionMany agree that online information seeking is not a straightforward procedure. Users of an online information retrieval (IR) system or search tool are heavily dependent on several factors when they search for relevant information: the right choice of keywords, knowledge of the behavior of the IR system, extensive use of browser tools or any other navigational aids, and some luck. A typical search session activity involves several stages: (a) query submission, where the user submits a query which (generally) represents his or her information goals, (b) invocation of the IR component, where given the input query, the search tools return a list of filtered documents for user inspection, (c) selection and navigation, where the user selects documents from the result page list for further inspections, or (d) reformulation, where the user submits a revised query which may be related to the initial query, and then restarts the process.When a user interacts with a search system, two important pieces of information may be recorded: (a) the query terms(s) issued and (b) the list of links or navigation trails followed from the results page. Query terms may be collectively interpreted as the informational goal users set out to achieve, while a navigation trail (or a list of links followed) may be interpreted as one of the processes undertaken to achieve this goal. Although users' information searching behavior differs from one user to another, in general this process can be summarized as a "mini-session" of query submissions intertwined with navigation trails. The total collection of mini-sessions of query submissions intertwined with navi...

Moving towards Adaptive Search in Digital Libraries

“…Furthermore, despite the risk of offering irrelevant suggestions in a system-guided search system, users might prefer having them rather than not [49]. On the other hand, it has also been shown that users are more inclined to submit new queries or resubmit modified queries than to navigate from the result set in a search environment that supports search and navigation [35]. Perhaps the best evidence for the usefulness of interactive search systems is the fact that even the big Web search engines have recently added more and more interactive features, e.g., Google's Wonderwheel 3 .…”

Online Instructors As Thinking Advisors: A Model For Online Learner Adaptation

“…Most faculty teaching online recognize that working within the context of a web-based setting, whether it is basic information retrieval or more complex content processing, is not a simple, straightforward task (Mat-Hassan & Levene, 2005). When assisting students working in an online course, it is important to recognize that students have individual differences that reflect prior knowledge and skills, intellectual capacity, cognitive ability, and learning preferences (Graf, Liut, & Kinshuk, 2010).…”

“…Navigation behavior is the ability of the learner to retrieve, follow, and link multiple forms of information within a web-based environment (Sullivan, Gnesdilow, & Puntambekar, 2011). Successful navigation behavior is dependent upon knowledge and understanding of the information system (Mat-Hassan & Levene, 2005) to allow the individual to find needed information quickly and be able to return easily to previous pages or links. Individuals can possess a variety of navigation behaviors based upon their prior experience working with hypertext, comfort with technology, and even their reading comprehension skills (Sullivan et al, 2011).…”

“…The initial step of identifying learning styles provides a broad view of how each student will interpret course content, but it does not clearly indicate how the student will navigate the course. Although specific classifications of navigation behaviors exist (Bousbia et al, 2009), there are general practices that are beneficial to all individuals trying to navigate an online environment (Mat-Hassan & Levene, 2005). Online instructors can create a multicomponent "map" that provides direction for interacting with the course and establishes a clear path to move between various parts of the course.…”

Assigned tasks are not the same as self-chosen Web search tasks

“…The most common method is to do log file analysis. [5] [9] [8] Such transaction logs analysis is common practice throughout search engine engineering, Web log analysis is used to discover the patterns of queries, the relative frequencies of different kinds of queries, the variety of queries actually made and the overall characteristics of user behavior in the aggregate. Web logs analysis also informs search engines about what results are the most popular, information that is fed back into the ranking systems.…”