/sparql-query2vec

Some approach for Sparql queries proccesing to use in ML tasks.

Primary LanguageJava

Approaches to vectorize Sparql queries and their metadata for Machine Learning tasks.

Some baselines

  • Rhassan code for algebra queries representation.
  • Rhassan code for graph pattern queries representation based in kmedoids.
  • Queries representations using sets of features for DeepSet architecture(working on..)

Run

For run use that

System.out.println("Try with some of this parameters:");

 java -jar file.jar kmedoids /path/to/input.csv /path/to/output.csv /path/to/ids_time.csv #-of-centroids

For generate edit distances vectors

java -jar file.jar edit-distance /path/to/input.csv /path/to/output.csv /path/to/prefixes #-of-cores

Last number(4) is the number of cores or proccess to run in paralell.

For generate vectors for deepset architecture:

java -jar file.jar deepset-features /path/to/input.csv /path/to/output.csv tables,joins,predicates /path/to/prefixes [--cores=numOfCores] [--length=numOfTuples] [--output-delimiter=symbolToDelimitColumns]

Compilation

Execute to compile:

mvn clean package

In the generated target you will find a graph-edit-distance-1.0-SNAPSHOT.jar