Data-driven recommendations for exploratory query formulation
Fatemeh Nargesian · 2014
We consider the problem of helping users explore and understand large, high dimensional structured datasets. Such users may issue queries to find data entries of interest. Given the high dimensionality and lack of detailed knowledge of the data, these queries are very likely to return no answers. In this thesis, we propose a new data-driven framework that reformulates queries to help guide users to data entities of interest. Our approach uses the data itself, without any need of supplying a corpus of queries or a domain ontology. We discuss how our approach can be adapted to provide recommendations over semi-structured or unstructured data.