Toward scalable keyword search over structured data

The success of search engines has spawned substantial interest in keyword search over structured data (KWS-S) in order to accommodate users who cannot issue a formal structured query or are unaware of the underlying structure of the data. KWS-S systems accept keywords as input, however unlike search engines, instead of documents they return relationships between the user's keywords. While KWS-S solutions have been around for over a decade, their widespread adoption has been hindered by several roadblocks.Trying to address some of these roadblocks is the motivation for this dissertation. We observed that current KWS-S solutions have unpredictable performance. In particular, for certain queries it takes too long to produce answers, and for others the system may even fail to return (e.g., after exhausting memory). We argue that as today's users have been spoiled by the performance of Internet search engines. KWS-S systems should return whatever answers they can produce quickly and then provide users with options for exploring any portion of the answer space not covered by these answers. Our basic idea is to produce answers that can be generated quickly as in today's KWS-S systems, then to show query forms that characterize the unexplored portion of the answer space. We address the challenges in generating, searching and using query forms to to allows us to bypass the performance problems inherent to KWS-S without compromising query coverage. Another problem that we tackle is that of non-answers. We note that, in general, when analyzing data sets, users not only gain information from queries that return answers—they also get information from queries that return no answers. With this in mind, we propose a KWS-S system that also exposes non-answer queries to the user. Additionally our system uses sub-queries of the non-answer queries to provide some insight into the reasons for the non-answer queries.