Bahareh
Sarrafzadeh,
PhD
candidate
David
R.
Cheriton
School
of
Computer
Science
Information seeking is a fundamental component of many of the complex tasks presented to us, and is often conducted through interactions with automated search systems such as web search engines. Indeed, the ubiquity of Web search engines makes information so readily available that people now often turn to the Web for all manner of information seeking needs. Furthermore, as the range of online information seeking tasks grows, more complex and open-ended search activities have been identified. One type of complex search activity that is of increasing interest to researchers is exploratory search, where the goal involves “learning” or “investigating,” rather than simply “looking-up.”
Given the massive increase in information availability and the use of online search for tasks beyond simply looking-up, researchers have noted that it becomes increasingly challenging for users to effectively leverage the available online information for complex and open-ended search activities. One of the main limitations of the current document retrieval paradigm offered by modern search engines is that it provides a ranked list of documents as a response to the searcher’s query with no further support for locating and synthesizing relevant information. Therefore, the searcher is left to find and make sense of useful information in a massive information space that lacks any overview or conceptual organization.
This thesis explores the impact of alternative representations of search results on user behaviours and outcomes during exploratory search tasks. Our inquiry is inspired by the premise that exploratory search tasks require sensemaking, and that sensemaking involves constructing and interacting with representations of knowledge. As such, in order to provide the searchers with more support in performing exploratory activities, there is a need to move beyond the current document retrieval paradigm by extending the support for locating and externalizing semantic information from textual documents and by providing richer representations of the extracted information coupled with mechanisms for accessing and interacting with the information in ways that support exploration and sensemaking. This dissertation presents a series of discrete research endeavour to explore different aspects of providing information and presenting this information in ways that both extraction and assimilation of relevant information is supported.
We first address the problem of extracting information — that is more granular than documents — as a response to a user’s query by developing a novel information extraction system to represent documents as a series of entity-relationship tuples. Next, through a series of designing and evaluating alternative representations of search results, we examine at how this extracted information can be represented such that it extends the document-based search framework’s support for exploratory search tasks. Finally, we assess the ecological validity of this research by exploring error-prone representations of search results and how they impact a searcher’s ability to leverage our representations to perform exploratory search tasks.
Overall, this research contributes towards designing future search systems by providing insights into the efficacy of alternative representations of search results for supporting exploratory search activities, culminating in a novel hybrid representation called Hierarchical Knowledge Graphs (HKG). To this end we propose and develop a framework that enables a reliable investigation of the impact of different representations and how they are perceived and utilized by information seekers.