Yet Another Knowledge Base

Ros Node for communication with a Apache Jena Fuseki triple store.

Requirements

Apache Jena Fuseki

Make sure you have the Java Developent Kit 11 installed:

java --version

If not, install it with

sudo apt install default-jre

Download: https://jena.apache.org/download/index.cgi

Run server from Fuseki root folder with:

.fuseki-server

This starts a Apache Jena Fuseki Server at localhost:3030.

If you get "Error: A JNI error has occurred, please check your installation and try again", you have sourced the wrong Java version and need to run

export JAVA_HOME=/usr/lib/jvm/java-11-openjdk-amd64

Once the server is running, open localhost:3030 in your favourite browser, navigate to "manage datasets" and add two new datasets, one called fs (Dataset type in-memory) as fact store, one called o (Dataset type Persistent).

Python

Go inside the yet_another_knowledge_base folder and run

pip install -r requirements.txt

Package information

The package contains a launch file, which starts the YAKoB_node, which is a ros wrapper for the knowledge base providing ros services for some knowledge base funcionality:

Messages

• Fact.msg:

  • string[3] fact: Triple of strings
  • string object_type: should be URI if the object of the fact is a URI eg. a concept inside an ontology. Or should be one of the ros datatypes, which will be mapped to an xsd:type for storage inside the triple store.

• QueryResult.msg:

  • uint32 n_results: Number of returned results
  • uint8 n_values: Number of values per result (example: for SELECT ?s ?p ?o n_values = 3)
  • string[] values: List of entities as strings returned from query
  • string[] types: States whether according value is of type url or literal

• YakobUpdateFact.srv: Service message for adding, deleting or modifying (not implemented) facts to the fact store.

  • Fact[] facts: Array of Facts to be updated
  • returns nothing

• YakobUpdateGraph.srv: Service message for adding a serialized graph to the fact store.

  • string graph: Serialized graph as string.
  • returns nothing

• YakobQuery.srv: Service message for querying the knowledge base with a SPARQL query.

  • string query: SPARQL query as single string
  • returns QueryResult result

Services

• yakob_add_ontology: add a Graph to the ontology dataset, uses YakobUpdateGraph.srv
• yakob_add_facts: add facts to the fact store, uses YakobUpdateFacts.srv. Subjects, predicats and objects that aren't part of an RDF namespace will be added to the internal namespace (rosparam /YAKoB/namespace)
• yakob_remove_facts: remove facts from the fact store, uses YakobUpdateFacts.srv
• yakob_query: query knowledge base with SPARQL query, uses YakobQuery.srv

Usage

After building the package and sourcing setup.bash you can run

roslaunch yet_another_knowledge_base start_yakob.launch

to start the knowledge base. Default namespace is the incorap namespace "http://dfki.de/incorap/incorap_domain#", which can be modified by changing /YAKoB/namespace in the launch file.

By default YAKoB is using Fuseki as triple store. To change to the rdflib implementation you need to change line 19 in YAKoB_node.py from

self.c = Connectors.FusekiConnector(self.datatype_map_path, self.namespace)

to

self.c = Connectors.RDFlibConnector(self.datatype_map_path, self.namespace)

yakob_playground.py is the node for your own experiments. Some hardcoded facts are in there, but feel free to go wild here.

Ontology

You find the incorap ontology in ressources/ontologies.