FAQ

• Q. What are first steps after building and running Fusepool Platform?
  A. You might want to do this:

1. Open a browser and point it to localhost:8080
2. Use ECS as described by it's readme and/or load some data using datalifecycle, such as from http://raw.fusepool.info/IREC/ (select one of the XML files)
3. You might need to do a reindex (http://localhost:8080/ecs/reindex) before the data shows up in firstswim

• Q. maven-gpg-plugin asks me to enter a GPG passphrase. What can I do?
  A. You probably need to deactivate the plugin by editing dice-linker/duke/pom.xml. See http://www.michaelpollmeier.com/maven-skip-pgp-sign-process/ for details.

• Q. How are implemented the Linking Engines (NamedEntityLinkingEngine, KeywordLinkingEngine, EntityhubLinkingEngine, EntityLinkingEngine) ?
  A. Ask question to Rupert (Luigi to provide more details, Nicolas will send him the question directly)

• Q. The new enhancers may need parameters for example to know in which triple collection to store the enhancements. The parameter should be passed through a HTTP POST with the data. It is not clear if and how this is possible.
  A. Create a new service using the archetype. Use the enhancer via OSGi as described on https://stanbol.apache.org/docs/trunk/components/enhancer/ and then you store in the triple collection as specified in the query argument.

• Q. What happens when more than one user tries to use an enhancer ? Where should be managed the concurrency ?
  A. Should not cause any problem. If you iterate over triples you should have a readlock

  final Lock readLock = mgraph.getLock().readLock();
  readLock.lock();
  try {
  //iterate over graph
  } finally {
  readLock.unlock();
  }

• Q. How to make Entities extracted from documents in the Contenthub and stored into a triple collection dereferenceable ?
  A. Good question. No dedicated stanbol side suport for this.It can be achieved by assigning URIs within the namespace of your service (e.g. http://localhost:8080/myservice/...) and having a method assigned to the wildcard as path handling these requests.

• Q. How does Stanbol differentiate between Data sets, Triple Collections, Graphs (NAMED GRAPH), and whatever else it uses to capture or logically store a collection of data?
  A. Dataset: about equivalent to a TcProvider: has no URI Triple collections have URIs Graphs are triple collections There is no way to differentiate from the URI if something is a graph or another resource (described in a graph). Check with TcManager if its a triple collection.

• Q. Can TDB store be directly Updated and Queried (bypassing Stanbol’s stuff)?
  A. With TcManager as in the archetype example you don’t use stanbol stuff (but are still independent of backend)

• Q. What’s the ETA on SPARQL 1.1?
  A. End of March

• Q. Is it possible to give more memory to TDB or whatever wrapper is used to access it, because the default is whatever memory is given for Stanbol (AFAIK). When I’m loading data, I want to dedicate a lot of memory, which may be different than the memory that is used to start the platform.
  A. Increase the memory of the VM used by the Stanbol instance.

• Q. Is it possible to export a triple collection with its index as a bundle that can be imported in an other Stanbol instance ?
  A. Serialize it in any RDF format, preferably N-TRIPLES.

• Q. Unit Test code for archetype ?
  A. Could add unit using Easymock. More relevant are high level tests as in http://svn.apache.org/viewvc/stanbol/trunk/integration-tests/ which show that the reqests are handled correctly.

• Q. How to query more than one graph ? It seems the sparql endpoint provided by Stanbol doesn’t support it.
  A. Use TcManager and/or raise issue against stanbol

• Q. Can someone write out a simple tutorial to make an existing REST-like web service work with Stanbol (in the quickest/easiest way)?
  A. This is with respect to the Enhancer; or, in fact, it’s about Stanbol in general. I think we really need at least an FAQ so that it will be easier to provide services that everyone can access. How do we access documents easily? How do we best provide services for everyone in Fusepool overall? Is it fine to say ok we have an active-learning based multi-label classification tool available via a web service? I’m working on an Enhancement Engine archetype, I could add a rest client.

• Q. What predicate URI should be used in the triples that are generated by the classification KMX enhancement engine?
  A. Use already defined properties when appropriate, such as ones from the w3 open annotation ontology or ones that are already present in stanbol.

• Q. KMX enhancement engine can give a classification score (confidence) in one of multiple categories. How to select which classifier should be applied? This might be the same as question number 2.
  A. Confidence: http://dev.iks-project.eu:8080/enhancer different enhancements have different scores. Property used: http://fise.iks-project.eu/ontology/confidence "0.3740056333976122"^^http://www.w3.org/2001/XMLSchema#double ; Use multiple enhancers, one for each type of classification.

• Q. Is there any testcode (or documentation) on how to insert entities or query them from entityhub?
  A. The archetype code contains example code on how to access EntityHub sites. Replace SiteManager with entityHub, the api is very similar.lFor adding entities via REST see http://dev.iks-project.eu:8080/entityhub