This project is part of my work done during my Google Summer of Code 2018 in collaboration with DBpedia
This repository contains the code needed to generate a CSV with the annotations on a language pair. Before running any code here, please, read the instructions to build a virtuoso endpoint and load the recommended triples and graphs.
The recommended IDE to run this project is Intellij IDEA.
This repository should include the .idea folder, to
ease the project importation to this IDE.
The recommended Java version to run it is Java 8.
This application will execute hundreds of queries against the SPARQL endpoint selected. So it is recommended to use a local endpoint to no overload original DBpedia server, as well as we also will need some additional triples that are not available by default on original endpoints.
This script will download and reify triples for the english and spanish DBpedia.
As we can observe by reading it, we need the following graphs for each language
- http://dbpedia.org/: contains the entities URIs, the same as the default DBpedia endpoint
- http://dbpedia.org/r: reified triples from tql releases on dbpedia.
- http://dbpedia.org/rml: The DBpedia mappings but transformed to rml format
We will use a different URIs for different languages. In the
example, if we use the spanish language, we would load
http://es.dbpedia.org/, http://es.dbpedia.org/r
and http://es.dbpedia.org/rml graphs
Additonally, the DBpedia ontology is needed to infer class hierarchy between properties and classes.
For the virtuoso configuration, you can use the virtuoso.ini file that exists inside this folder. It should not been used as a drop in replacement of installed .ini, but as a reference to fix problems, if any exist.
To execute the program and produce the needed annotations
CSV file, you only have to run the InconsistentMappings
class, starting on the main method.
The params to edit are the SPARQL endpoint and the RDF graphs to use with each language.
It may be needed to change the SPARQL endpoint constant also in other classes, like DBO class.
The work done by Nandana was almost ready for production, but it still worked weirdly on some executions and some attribute names were interchanged, so the CSV file wasn't consistent with the published attributes.