Recommender System for the SemaGrow Stack federation

The Recommender System is a piece of software - entirely based on JAVA - that computes meaningful combinations between some datasets federated by SemaGrow, generating a new triplestore: the “Recommender Database”. This work was funded by the European Commission under EU FP7 project SemaGrow (Grant No. 318497).

The Recommender System computes meaningful combinations between two or more datasets federated by SemaGrow: the computation of combinations is based on the matching of AGROVOC URIs between datasets.

System Requirements

• java >= 1.6 (mandatory)
• git >= 1.8.1.4 (to download the project from GitHub: other solutions may be adopted)
• maven >= 3.0.3 (to edit the code and build a new jar: you can also work with the provided command line application, without using Maven)
• linux environment (the provided command line application comes with a bash script to execute the code. A developer can replace the bash script with another one, as a bat script for Windows)

Execute the command line application

The folder executable contains the command line application, including the bash script start.sh to run the recommender system (note: if the bash can't interpret the script, try to run the dos2unix start.sh command).

The command line application is composed of some folders:

• bin: containing the compiled JAVA classes
• lib: containing the needed JAR files (all dependances can be found in the maven-source/target/classes/ folder)
• resources: containing the configuration file defaults.properties (that can be found in the maven-source/target/classes/ folder)

The file defaults.properties contains input parameters for the recommender system:

• sourceFilePath is the path of the input file, containing one URI for each line. The system computes recommendations for each URI available in this file (e.g. /work/recommender/recomm/input.txt)
• outputFilePath defines the location of the output files. The path should include the default name of the file. Then, the system will add a timestamp to each produced file (e.g. /work/recommender/recomm/data/output.xml)
• sparqlEndpointSG is the SemaGrow SPARQL endpoint federating target datasets of interest. It should contain at least the dataset of URIs defined in sourceFilePath and the output datasets (whose entities will be recommended to the client) specified by the target_rdftype parameter
• max_recommendations is the maximun number of desired recommendations for each entity
• target_rdftype defines the output of the recommender system: only concepts of that type will be considered recommendation for a specific URI

The Recommender System can run also by querying two individual SPARQL endpoints (for more than two endpoints, there is the need to use the federated mode). In the file defaults.properties simply configure the following properties:

• type_recommendation should be set to individual
• sparqlEndpoint1 is the SPARQL endpoint of the dataset of URIs defined in sourceFilePath
• sparqlEndpoint2 is the SPARQL endpoint of the output datasets (whose entities will be recommended to the client) specified by the target_rdftype parameter

The Maven project

The folder maven-source contains the Maven source project. You can use it to edit the code or to build the "recommender system" distribution zip file (i.e. the command line application):

cd maven-source/
mvn clean install

You will find the created zip file in the directory maven-source/target. In addition to that, the directory maven-source/target/classes/ contains also all dependances (jars to be added to the classpath of the command line application) and the configuration file default.properties (to be added to the classpath of the command line application).

License

Creative Commons Attribution 4.0 International (CC BY 4.0). See Legal Code