LPRules: Rule Induction in Knowledge Graphs Using Linear Programming

This repository contains the code and the data used for the experiments in the paper "Rule Induction in Knowledge Graphs Using Linear Programming" by Sanjeeb Dash and Joao Goncalves, AAAI-23.

Requirements:

• The code was tested only on Linux.
• The code is written in C++ and requires a C++ compiler.
• The code uses the commercial solver IBM ILOG CPLEX.

Folders description:

• code: contains the code and the Makefile.
• data: contains the 5 datasets used in the paper.
• runs: contains the parameters files and scripts to run the 5 datasets.

How to install the code:

• Install IBM ILOG CPLEX.
• Edit the file Makefile in the directory code and add the paths to the directory cplex and concert in the lines:
  • CPLEXDIR = path_to_cplex/cplex
  • CONCERTDIR = path_to_cplex/concert
• In the directory code type make.

How to run the code without using previous known rules:

The following instructions show how to run LPRules with the dataset UMLS.

• Go to the directory run/UMLS.
• Execute the command: ./run_lprules_in_parallel.sh p_UMLS.txt outUMLS 46 where p_UMLS.txt is the parameter file residing in the directory run/UMLS, outUMLS is the name to be used in output files and 46 is the number of relations in the UMLS dataset.
• The results are presented at the end of the file results_outUMLS.txt.

The number of relations in each dataset is: UMLS 46, Kinship 25, WN18RR 11, FB15k-237 237, YAGO3-10 37.

How to run the code using previous known rules:

The following instructions show how to run LPRules with the dataset UMLS assuming the existence of a set of rules in the file input_rules.txt.

• Go to the directory run/UMLS.
• Edit the file p_UMLS.txt and set the parameter run_mode to either 1, 2, or 3. The meaning of each of these values is: 1 - scenario B in the paper, 2 - scenario C in the paper, 3 - scenario D in the paper.
• Execute the command: ./run_lprules_in_parallel_read_rules.sh p_UMLS.txt outUMLS 46 input_rules.txt where p_UMLS.txt is the parameter file residing in the directory run/UMLS, outUMLS is the name to be used in output files and 46 is the number of relations in the UMLS dataset.
• The results are presented at the end of the file results_outUMLS.txt.