I'm doing a project to transform the Relational Database to RDF database to support the semantic web and ontology.
I found many approached for doing that, and some tools. For who have an experience in that topic.
My question is: what are the current improvements (or gaps) that I can do to improve the transformation process of Relational database to RDF (like transform additional constraints, or store RDF as graph to increase the query performance, or improve the data integration process between relational databases etc)???
asked 31 Jan '13, 10:29
I'd like to see a tool that converts relational data (JDBC, ODBC, etc.) to into a correct and compatible N-Triples file(s) that I can feed into triple stores, map/reduce pipelines, etc. The basic mapping should be compatible with
MOST IMPORTANT: I've seen open source tools that do this, but they create data structures in RAM which grow along with the size of the data set and eventually cause memory exhaustion. The system must process data in a streaming mode and the map/reduce architecture works so well for that.
If you built something like this on top of Hadoop you could certainly build a cluster that could suck up a relational database of the size of the Pentagon's.
This is based on a faith that RDF is the "universal solvent", that it is the best data integration tool, that a combination of databases and logic, rules and queries. We can write rules in the RDF world to match up the hash that comes out of the direct mapping to our point-of-view vocabulary.
answered 31 Jan '13, 12:39
Why do you need to transform it? You can keep the relational database and have it generate triples. Just like many millions of websites generate HTML out of their relational database content, you can do the same with RDF. After all, RDF is a resource description, and it doesn't mean you need an RDF database to generate RDF.
answered 31 Jan '13, 17:00
Be sure to look into existing open source tools so you can what others have done. For instance, check out this question: "R2RML processor for PostgreSQL databases?"
Also, search this site for R2RML to see what you can learn about it. You'll find other exisitng tools such as Spyder.
answered 31 Jan '13, 21:55