A platform for implementing and benchmarking defeasible reasoning algorithms, including current approaches for rational and lexicographic closure and new model-based implementations for rational closure, Lehmann's lexicographic closure, and Casini et al.'s count-based lexicographic closure.
- Maven 3.3+
- Java 15+
To build the project, run:
mvn package
Note: Knowledge bases should be placed in the data/ directory.
To run queries on knowledge bases using all the built-in reasoning services, excute the following command:
java -cp target/MBDR-1.0.0-jar-with-dependencies.jar com.mbdr.App <filename> <query>
E.g.
java -cp target/MBDR-1.0.0-jar-with-dependencies.jar com.mbdr.App "penguins.txt" "p|~f"
To run benchmarks for the implemented algorithms, run the following commands:
Construction:
java -cp target/MBDR-1.0.0-jar-with-dependencies.jar com.mbdr.benchmarking.BenchmarkConstruction
Entailment:
java -cp target/MBDR-1.0.0-jar-with-dependencies.jar com.mbdr.benchmarking.BenchmarkEntailment
To run the Javalin web application:
java -cp target/MBDR-1.0.0-jar-with-dependencies.jar com.mbdr.web.WebApp
Knowledge base generation requires the latest version of scala.
The benchmarking data set was generated using the Knowledge Base Generation tool from SCADR (2021). It can be generated using KBGenerationScript.sh in the scripts/knowledge_base_generation/ directory.
Scripts for graphing construction and entailment results are included in the scripts/graphing/ directory.
Install the required modules before running:
pip install -r requirement.txt
To generate the javadocs, run the following command in a Linux terminal/shell:
javadoc -sourcepath ./src/main/java/com/mbdr -d ./docs -classpath target/MBDR-1.0.0-jar-with-dependencies.jar $(find ./src/main/java/com/mbdr/ -name *.java)
Visit the project website: https://twiddleproject.com, for more information.
- Jaron Cohen
- Carl Combrinck