From STLab

Saarbrucken Developers Workshop 2011

Contents 1 Saarbrucken Developers Workshop 2011 1.1 Day 1 - 05/04/2011 1.2 Day 2 - 06/04/2011 1.3 Day 3 - 07/04/2011

Day 1 - 05/04/2011

Today each group has synthetically described the status of the work done so far.

Here the main points of the discussion:

• The necessity of more modularity for openNLP and so also a multilingual functionality of this componed.
• The output of the enhancer to be used in the entityuhb with adding of sameAs relations. This might require KReS functionalities in the Entityhub. The KReS-Entityhub integration will start during next days of the workshop.
• The storage is accessible via the TcManager. By the way the TcManager is a additional layer over the physical storage, that can be potentially any triple store (TDB, SDB, Sesame, etc). The storage developed by turkish guys is able to manage both T-Box and A-Box.
• Archive with rules in the configuration file. The key idea here is to provide easy functionalities in order to configure Stanbol with a set of rule (mainly for the refactor). A default set of rules should be provided.
• The ontology generator component is the second component developed by turkish guys. It provides API for the generation of OWL ontologies starting from CMIS and JCR. The ontology generation is performed by the classical idea of "bridges" based mapping. It means that anything in the original source that has to become a class is mapped to a class via a class-bridge, a property via a property-bridge and so on. The transformation is a data-exchange as it is based on a RDF/OWL dump of the original source. It has been well accepted by other guys the idea to adopt D2R-like mechanisms for graph virtualization and data access via SPARQL. The ontology generator provides also API, services and HTML interfaces which allow to modify ontologies. Has been said the hypothesis to translate SPARQL to CMISQ and SQL query. In this way it is thought that the query would be more efficient.
• The ontology generator has a lot of common points with Semion, so an integration or better a fusion is not a bad idea.
• A good point for KReS has been suggested by Ruperth. It consists of a client-side module for refactoring. It can be realized by sending the rule in JSON format to a client application based on RDFquery which embeds the core refactoring mechanisms of KReS.
• The presentation of VIE by which a text documents can be composed with objects. Each object is editable and is part of a RDF graphs and nodes. It has an indication process similar to scout.
• The benchmark quality test for FISE enhancers. Here is set a list of desired (and undesired) known outputs function of particular known inputs. So the test verify if given that particular input there are effectively the expected outputs.

Moreover, has been presented also some demo related to vie:

• Intelligent project controlling tools: an example of project reporting and monitoring that is composed by editable objects those are part of a RDF graph.
• A chat (palsuapp.info) that as demo before as editable objects belong to an RDF graph.

Day 2 - 06/04/2011

Today we have focused our work on trying to integrate two Stanbol components: the PersistenceStore (PS) and the EntityHub (EH). Key points are the following:

• The PS provides API and RESTful services which allow to manipulate the OWL model in a persistent way via Clerezza. We have found a strange thing in the PS about how ontologies and and axioms are passed to the PS components. In fact, they are always passed as parameters to the methods of the PS. An issue has been open on JIRA (STANBOL-154).
• We have first integrated the PS store in OntoNet (see STANBOL-145), but that has introduced a strong and unnecessary dependency between OntoNet and the PS. For that reason (see comments related to STANBOL-145) the integration has been removed.
• We have had a discussion with turkish guys about an improvement that from our could be interesting. It concerns the implementation of a concrete OWLOntologyManager that grounds on the PS in order to have a persistent layer (somehow native) for the OWL API. No issue has been opened as in next days both us and the turkish guys will first investigate on the complexity and the feasibility of that improvement.
• We have decided to use integrity check demo for testing the integration between former called KReS components and the PS. The idea is to use the PS in order to modify integrity constraints.
• We have had a discussion with Ruperth about basically how to use both from API and RESTful side the EH. He also gave us a current index of DBpedia index that might be used in our installation of Stanbol.
• Trying to use the EH from API we have found a verbose procedure when you want to fetch the RDF graph related to an entity. It depends on the fact the EH API are much more oriented to the index (Lucene/Solr) then to RDF itself.
• The EH can be integrated in all our demos. We will start from the dulcifier that from our point is much easier as it already provides a GUI.
• Both OntoNet and PS exposes OWL API and OWL-Link bundles. That must be solved ASAP just because when both components are deployed some unsatisfied constraints may occur.

Day 3 - 07/04/2011

Today we have focused our work on trying to integrate two Stanbol components: the PersistenceStore (PS) and the EntityHub (EH). Key points are the following:

• The PS can be used to manage imported ontologies inside the scope. In this way the user can add or remove classes, properties, individuals, etc... The problem is that the PS actually cannot manage imports, so we have asked to Chian to add functionalities for imports. We have check the PS functionality by installing all the component in the bundle (to activate all the bundles we need to remove our owl-link API and the PS owlapi) and running a curl request to add scope. Then the core and custom scope ontologies can be seen from the PS. Therefore PS and the ONManager have the sames access to Clarezza's space.
• N.B. For the integrity-check it is important that the ontologies can be managed both in memory and also via PS. So the change in memory will reflect change in the PS and vice versa. This way the ontologies changed via PS must be seen also from ONManager and vice versa.
• The EH has been integrated with the dulcifier engine. Now the entities (person, place, etc) are retrieved by Entityhub and not via Internet as before.
• In the meantime we have started to align the integrity-check demo with the new Stanbol code.