Minutes of LarKC WP4 Teleconference, 11.00 -12.00, March 6, 2009
The skype ID of the chair: larkc-wp4
Participants
Barry(STI), Vassil (?OntoText), Neso(?OntoText), Blaz (Cyceuro), Luka (Cyceuro), Daniele Braga (CEFRIEL), Daniel Dell'Aglio (CEFRIEL), Axel (Stuttgart), Alex (Stuttgart), Jose(MPG), Lael (MPG), Hans (MPG), Yi Zeng(WICI), Ruizhi Wang (WICI), Gaston (VUA), and Zhisheng(VUA)
Agenda
- Month18 Deliverable D4.7.1 Initial Evaluation and Revision of Plug-ins deployed in use-cases
- Month18 Deliverable D4.3.1 strategies and Design for Interleaving Reasoning and Selection of axioms.
- How to use Latex and SVN for LarKC deliverables
- Questions on Reasoner plugins and LarKC API
- Date for Next WP4 teleconference
D4.7.1: Initial Evaluation & Revision of plug-ins deployed in use-cases
due M18
Person-month commitment: Innsbruck: 1 PM, ?CycEuro: 2 PM, HLRS: 2 PM, ?OntoText: 3PM, VUA: 5PM, WICI: 1PM.
TODO: look at the evaluation measures defined in deliverables from WP6,7a,7b, and to translate these to evaluation measures for the reasoners
TODO: find out which reasoner plugins they intend to use in the use-case
Make sure that there is a link with the D1.4.1 deliverable on methodology of evaluation
Workspace for D4.7.1: http://wiki.larkc.eu/LarkcProject/WP4/D4.7.1
Contributions by each partner:
UIBK: Will contribute a chapter on evaluation of Rule-based reasoning
CycEur: Evaluation of larkc deciders
HLRS: Evaluation of parellel reasoning
VUA: Contribute chapter evaluation of standard reasoners and approximate reasoning
Ontotext: Evaluate the use-case WP7a/b req. for the plugins
WICI: Evaluation of granular reasoner (Evaluation of case-based reasoners may be considered)
MPG:: will contribute some work which is based on the work for WP1 and WP2.
CEFRIEL:: will look at the evaluation measures of the case study of Urban computing
Plan an evaluation meeting which covers the work of WP1, WP2, WP4, and WP5, and case studies in the LarKC Milan meeting in May 2009.
D4.3.1: Strategies & Design for interleaving reasoning and selection of axioms
due M18
Plan: use and extend existing work from PION and Stefan's work
PION = select consistent subset of axioms, reason, repeat
Stefan = select subset of ontology, classify, repeat
TODO: implement selection and reasoning as two different plugins,
- both for PION and for "Stefan"
TODO: think of other selection strategies for PION (also for "Stefan"),
- for example
- the activation spreading selector from ?OntoText - the visualisation/human-selector from WICI - graph-measures selection from MPG (in-betweenness, centrality)
Workspace for D4.3.1: http://wiki.larkc.eu/LarkcProject/WP4/D4.3.1
Ontotext:: for the work with the activation spreading selector,
delay the discussion after Month 12. Will discuss it in next meeting, and discuss it with someone from Ontotext who will attend the WP2 plugin workshop.
WICI:: will contribute the work of user-profile-based selection and granular reasoning.
MPG:: will explore it further. The student who will visit MPG from Amsterdam will contribute to it.
How to use Latex and SVN for LarKC deliverables
See the svn wiki page:
http://wiki.larkc.eu/LarkcProject/HowToSvn
Lael: how to track changes in latex
Gaston: 1 possibility for tracking changes in Latex: http://trackchanges.sourceforge.net/
Questions about reasoning plug-ins and the LarKC API
Take a look at the example given in section "How to pass ?RdfGraph by ?DataSet/TripleSet reference" of the Data Layer API's wiki page (http://wiki.larkc.eu/LarkcProject/WP5/DataLayerAPI).
Supposse that we create an eu.larkc.core.data.RdfGRaph object (line: ?RdfGraph graph = df.createDataSetGraph(uri, ds);) after this the next logical step could be to call a reasoner plug-in and pass it that ?RdfGraph that we just created so it can perform the resoning task. The reasoner has only one way to get the triples out of this graph (as far as we understand) which is by means of the method getStatements() in the class eu.larkc.core.data.RdfGRaph, which returns an Iterator. Suppose that the reasoner is only a wrapper of an existing reasoner (e.g. Pellet, RAcer, etc.) and thus it needs to pass the triples to the other reasoner. Currently, the only way to do this is by reading the statements 1 by 1 and building a new rdf graph that will be passed to the reasoner being wrapped. But this is very inefficient (imaging processing billions of triples 1 by 1 to creat a new graph). Is there any way to extract the triples from the eu.larkc.core.data.RdfGRaph using a URL. In that case we would only have to pass the URL of this graph to pellet, Racer, etc. We thought about using the method getName() in the class eu.larkc.core.data.RdfGRaph but of course this only works when the returned URI is actually an URL and assuming the graph is behind that URL.
Date for Next Monthly WP4 Teleconference
17 April, 2009, Friday, starting at 11.00 (CET) (The second Friday of every month by default)