UMBC ebiquity
LinkData service helps produce RDF linked data from tables

LinkData service helps produce RDF linked data from tables

Tim Finin, 9:59am 7 March 2012

Link Data is a nicely done Web site to help people produce RDF data from simple excel spreadsheets. It appears to be the work of researchers at the RIKEN BASE group at the RIKEN Yokohama Institute in Japan. The approach is straightforward and consists of three steps: creating a template, downloading it as an excel spreadsheet and adding your data, and uploading the result to the site for conversion and publishing.

In the first step, you use the site to create a template for your table, each row of which will be mapped to RDF data about a single subject. The first column of a row must represent the subject and the remaining columns its properties. After specifying the number of columns, you enter into each a string or a URI representing the property. If you enter a string (e.g., ’employer’), the system shows some suggested URIs drawn from the OBO ontologies that you can select instead of the string. You can also specify the cell values will be literals of type date, time, integer or float.

After downloading your spreadsheet template to your computer, you will see that the metadata is embedded in the initial rows of the table. Your next task is to enter your data, either as strings or URIs, as appropriate.

The final step is to upload the spreadsheet with your data to linkData.org, provide some additional data, and have it converted to RDF and make available on the site. Along the way you can see the results via the W3C validator as serialized in RDF/XML or depicted as a graph. You can also see how your data is connected to datasets in the LOD cloud.

Here’s the result of a simple test, in which I created a data set about people, their employers and their countries of residence.

The approach is simple and has many limitations, but I liked the Web interface and workflow. We’ve done some work in this space with RDF123 and are currently working on automating the process of producing five-start linked data.


Comments are closed.