| Abstract |
In my previous WAI talk, where I motivated the advantages and opportunities of recognizing emergent schemas of the semantic web, which are (a) faster query processing, (b) the ability to use relational tools on semantic data and (c) better user understanding of this data. The "emergent schema" is the rough structure that one can find in the billions of triples in the semantic web, even if these do not have a declared schema, or only a partial one, or if the schema is emerging from grass-roots initiatives.
This time, I will describe *how* this can be done. The task is two-fold: (1) to recognize an as small as possible relational schema that "covers" an as large percentage of the triples as possible by looking at the data instances, and (2) to give the recognized tables, columns and relationships names, that are understandable for users. I will discuss the results of my students Pham Minh Duc and Linnea Passing on this topic in creating the MonetDB RDF support module, which revolves around recognizing and exploiting this emergent schema. |