The Data Game

The data relevant to a query or analysis task is usually stored in various data sources, therefore, users often have to integrate information from several data sources. This is challenging as each data source may represent information in a distinct form, e.g., each data source may refer to the same entity under a distinct name. Users have to translate their queries to forms that are understand- able by underlying data sources. This process is traditionally done by writing a set of potentially declarative rules called mappings, which takes the query or data organized in one form and translate it to the query/ data under another representation. It, however, takes a very long time, a great deal of manual labor, and constant expert attention to develop and maintain mappings. One may use supervised learning techniques to develop mappings. How- ever, training data is hard to find for data integration. As the underlying data sources frequently evolve, one has to repeatedly find fresh training data to re-train mappings. Thus, mapping development and maintenance remains and is becoming ever more challenging in the face of rapidly growing number of available data sources. This work focuses on improving interaction with databases through the medium of playing a game. Interaction can either take place between the user and the database system or between multiple database systems. Overall, our goal is to improve the satisfaction of the user such that they get the information desired, whether it is through improving interaction with a single database or retrieving additional information from other databases. Quantifying whether these interactions were successful or not is based entirely on the setting. Between the user and the database system, improving interactions entails providing the results for which the user was searching. When considering the interaction between two database systems, success is determined by providing the correct results, but also by how much new information is gathered by the database system.

Co Advisor: Arash Termehchy
Co Advisor: Eduardo Cotilla-Sanchez
Committee: Alan Fern
Committee: David Maier
Committee: Liang Huang
GCR: Kyle Niemeyer