Graph-Factorization

Graph Factorization project creates compact representations of RDF graphs by reducing the number of frequent star patterns (FSP) in the RDF graphs. In factorized representations of RDF graphs the number of RDF triples is reduced while preserving the knowledge encoded in the data. Graph factorization project requires a class and a set of corresponding properties, involved in frequent star patterns, to perform factorization. EFSP and GFSP perform FSP detection by identifying classes and properties involved in frequent star patterns. The source code of factorization, EFSP, and GFSP are provided along with an example of sensor data containing temperature observations from LinkedSensorData. Files required to run the example are provided in /GraphFactorization/database folder.

Running this project requires following steps:

1. Configuration
2. FSP Detection
3. Factorization

1. Configuration

In this step a docker image of ubuntu is created, and ubuntu image along with an other docker container running virtuoso endpoint are started using docker-compose.

$cd Graph-Factorization
$docker build -t ubuntu1804 .
$docker-compose up -d
$docker exec -it graphFactorization bash

2. FSP Detection

Two approached EFSP and GFSP are provided to detect frequent star patterns. Each approach is explined below long with its usage.

2.1 EFSP

EFSP performs an exhaustive search over the frequent subgraphs identified by a graph mining algorithm. We have implemented EFSP, which uses gSpan as baseline. Running EFSP involves following steps:

Step 1. Running RDF2Graph

gSpan requires a graph format with vertices(v) and edges(e), therefore, RDF2Graph converts RDF graphs to the required graph format. Given below is the command to run RDF2Graph.

$python3 main.py $path_to_RDF $RDF_format $path_to_class_property_file $path_to_ve_graph_folder $path_to_folder_to_save_exe_parameters

Parameters

$path_to_RDF - It is the path to folder containing RDF data.

$RDF_format - It is a strinng mentioning the input format of RDf data.

$path_to_class_property_file - Path to the file containg a class and the properties of the class to be checked for FSP.

$path_to_ve_graph_folder - Path to a output folder to save the ve graph format of the RDF data.

$path_to_folder_to_save_exe_parameters - Path to a folder to save the execution time and other parameters observed during execution.

Running RDF2Graph using Example

$cd RDF2Graph
$python3 main.py /data/database/RDFOriginal/ n3 /data/database/classProperties/Temperatureproperties /data/database/veGraph/  /data/database/Result/

Running using shell script.

$sh scripts/runRDF2Graph.sh /data/database/db/rdf2graph /data/database/veGraph/ /data/database/Result/

Step 2. Running EFSP

Command to run EFSP

$python3 main.py -s 1 -d True $path_to_veGraph_file $path_to_folder_to_save_freq_star_patterns $path_to_folder_to_save_exe_parameters $path_to_file_containing_total_class_instances $path_to_class_property_file $path_to_folder_to_store_identified_class_properties

Parameters

$path_to_veGraph_file - Path to file containing ve graph representations.

$path_to_folder_to_save_freq_star_patterns - Path to folder to save frequent star patterns

$path_to_folder_to_save_exe_parameters - Path to a folder to save the execution time and other parameters observed during execution.

$path_to_file_containing_total_class_instances - Path to file containing total class instances.

$path_to_class_property_file - Path to the file containg a class and the properties of the class to be checked for FSP.

$path_to_folder_to_store_identified_class_properties - Path to folder to store class and properties, in json format, identified by EFSP.

Running EFSP using Example

$cd EFSP
$python3 main.py -s 1 -d True /data/database/veGraph/TemperatureObservationgraph.data /data/database/frequentStars/ /data/database/Result /data/database/instances/TemperatureObservation /data/database/classProperties/Temperatureproperties /data/database/json-file/

Running using shell script.

$sh scripts/runEFSP.sh /data/database/db/efsp /data/database/frequentStars/ /data/database/Result/ /data/database/json-file/

2.2 GFSP

GFSP adopts a Greedy approach to identify properties involved in frequent star patterns. Running GFSP involves following steps:

Step 1. Running GroupBy

GFSP requires a list of star patterns in RDF graphs, and corresponding multiplicities, therefore, GroupBy extracts lists of stars by running GroupBy queries over a virtuoso endpoint containing RDF data. Given below is the command to run GroupBy.

$python3 main.py $sparql_endpoint $path_to_class_property_file $path_to_folder_to_store_starList $path_to_folder_to_store_total_class_instances $path_to_folder_to_save_exe_parameters

Parameters

$sparql_endpoint - SPARQL endoint URL

$path_to_class_property_file - Path to the file containg a class and the properties of the class to be checked for star patterns.

$path_to_folder_to_store_starList - Path to folder to store star lists.

$path_to_folder_to_store_total_class_instances - Path to folder to store total class instances.

$path_to_folder_to_save_exe_parameters - Path to a folder to save the execution time and other parameters observed during execution.

Running GroupBy using Example

$cd GroupBy
$python3 main.py http://virtuosoGraphF:8890/sparql /data/database/classProperties/Temperatureproperties /data/database/starList /data/database/instances /data/database/Result/

Running using shell script.

sh scripts/runGroupBy.sh /data/database/db/groupBy /data/database/starList/ /data/database/instances/ /data/database/Result/

Step 2. Running GFSP

Command to run GFSP

python3 main.py $path_to_file_containing_starLists $path_to_class_property_file $path_to_folder_to_save_freq_star_patterns $path_to_folder_to_save_exe_parameters $path_to_file_containing_total_class_instances $path_to_folder_to_store_identified_class_properties -s 1

Parameters

$path_to_file_containing_starLists - Path to file containing star lists.

$path_to_class_property_file - Path to the file containg a class and the properties of the class to be checked for FSP.

$path_to_folder_to_save_freq_star_patterns - Path to folder to save frequent star patterns

$path_to_folder_to_save_exe_parameters - Path to a folder to save the execution time and other parameters observed during execution.

$path_to_file_containing_total_class_instances - Path to file containing total class instances.

$path_to_folder_to_store_identified_class_properties - Path to folder to store class and properties, in json format, identified by GFSP.

Running GFSP using Example

$cd GFSP
$python3 main.py /data/database/starList/TemperatureObservation /data/database/classProperties/Temperatureproperties /data/database/frequentStars /data/database/Result /data/database/instances/TemperatureObservation /data/database/json-file/ -s 1

Running using shell script.

$sh scripts/runGFSP.sh /data/database/db/gfsp /data/database/frequentStars/  /data/database/Result/ /data/database/json-file/

3. Factorization

Create Maven Package

$mvn package

Run Factorization

$java -jar target/RDFFactorization-0.0.1-SNAPSHOT.jar $path_to_json_file $path_to_input_RDF $path_to_output_RDF $RDF_format

Parameters

$path_to_json_file - Path to json file containing a class and corresponding properties to perform factorization. An example of json file is provided in database/json-file folder.

$path_to_input_RDF - Path to the folder conataining original RDF to be factorized. An example of original RDF data is provided in database/input folder.

$path_to_output_RDF - Path to the folder to save the factorized RDF data.

$RDF_format - Format of RDF files.

Running Factorization using Example

Example json file and original RDF data are provided in the database folder. Use the command below to run the example.

$cd RDFFactorization
$java -jar target/RDFFactorization-0.0.1-SNAPSHOT.jar /data/database/json-file/TemperatureObservation /data/database/RDFOriginal/ /data/database/RDFFactorized/ NTriples