Dissertation

Finding and accessing existing, relevant records in a timely fashion is often prohibitively difficult. Traditionally, we are forced to hope that such an investigation has already been done, and that we can find an existing answer to our question through traditional search. Continuing with the Olympics example, we can find a Web blog with a relevant result as shown in Figure 1, but our curiosity is held hostage to the static materials that conform to the author’s intent, not ours.

If these intents differ even in the slightest, the absence of supporting materials for inspection or reuse leave us stranded with only the author’s conclusions. Unfortunately, this is overwhelmingly the case with traditional analytical and publishing methods. Critical connections exist, but are left implicit and thus un-exploitable in situations where an analyst tries to leverage a previous analyst’s relevant materials.

We hypothesize that analysts within disparate contexts and environments can establish the explicit connections that enable other analysts within independent contexts to perform subsequent investigations faster and more comprehensively, and the objective of this thesis is to design and implement a framework that counteracts the omissions of connection that occur throughout and among analytical investigations to test our hypothesis.

Few would disagree with the Web’s ability to foster connections; it has billions of pages and at least a trillion links. So, it is a natural choice as an environment to establish explicit connections among analysts’ materials. The Linked Data community has made important progress in reapplying Web principles to shift data representations from being simply “on the Web” to being natively “in the Web”.

But, more work needs to be done from this perspective to broaden its scope from just “data in the Web” to a more complete “analytics in the Web”. Meanwhile, the Visual Analytics community has made important progress across the broad spectrum necessary to help analysts gain insights and make decisions, from understanding human cognition to building infrastructure that includes shifts towards “analytics on the Web”. But, more work needs to be done from this perspective to fulfill the potential offered by a more complete and interconnected “analytics in the Web”.

Computation lies at the heart of analytics, and reusing others’ data and algorithms can significantly reduce the time and effort it takes to accomplish an investigation. But, when analysts reuse prior materials, they are fundamentally recontextualizing those materials into new frames to suit their own objectives. This dynamic is an inherent part of deriving information, but at the same time presents challenges to consumers with their own, unique objectives that also bring their own biases.

On the other hand, if this dynamic can be adequately captured, it presents a significant opportunity for discovery and reuse that spans analytical efforts. This thesis offers the following contributions to address challenges caused by a dearth of analytical connection, so that subsequent investigations can be done more quickly and more comprehensively.The main contributions are: 1. A definition of Subsequent Analyst to frame the challenges faced when situating within disparate, unanticipated analytical contexts. 2.

Identified challenges that Linked Data publishers face by conducting information gathering interviews with more than 100 publishers. 3. Design and implementation (DataFAQs) of a usage-centric data quality assessment framework to empower the data consumer and inform the data publisher. 4. (a) Case study of applying Linked Data methods to perform Visual Analytics.

(b) Application of DataFAQs to assess 4a results for two quality aspects. (c) Evaluation of performance metrics using the results of this framework compared with results from the state-of-the-art VAST Challenge.