Facilitating the discovery of open government datasets through an exploratory data search interface

The primary area of investigation for this paper is the process of open data discovery, specifically, how novice users search for open government datasets and how this process can be improved. The problem of search on open data portals has featured consistently in Canadian Open Government consultations. A literature review of open data initiatives and processes reveals that open data search is a browsing or investigative task rather than a factual lookup or subject search task. Researchers have proposed exploratory search interfaces as alternatives to traditional search interfaces to support learning and investigative tasks in the information retrieval domain. This paper elaborates on a study conducted to evaluate the usefulness of a visualization-based exploratory data search interface to help novice users discover relevant datasets. A special feature of this interface is the use of variable names in addition to dataset description for search. Today, data search systems on open data portals tend to rely on the text contained in metadata and dataset descriptions to facilitate keyword search. However, the variable names contain important information about the content and structure of a file. This study also probed the role that variable names could play in search, specifically as a means of facilitating exploration and relevance assessment.