Draft Minutes of ICOM TC Meeting, June 24, 2009

Minutes of ICOM TC Meeting, June 24, taken by Eric S. Chan

Agenda:

1. Roll Call

2. Approve draft minute from June 10 TC Meeting

3. Continue discussions to reconcile Beehive object model and SIOC ontology.

4. Continue discussions of JPA and jenabean prototype for ICOM

5. AOB

1. The following eligible members were present

Siegfried Handschuh

Patrick Durusau

Ramesh Vasudevan

Eric Chan

Siegfried introduced Laura Dragan, who participated in this meeting. Eric is providing information to Laura for joining the ICOM TC either as an observer or a member.

2. Approval of the draft minute from June 10 Meeting was deferred.

3. Discussions to reconcile Beehive object model and SIOC ontology

Eric initiated the discussion of sioc:has_container and sioc:has_parent properties in SIOC ontology. The difference between these two properties is the domain: domain of sioc:has_container is sioc:item and domain of sioc:has_parent is sioc:container, see http://rdfs.org/sioc/spec. SIOC specification does not specify the cardinality for these two properties, since SIOC is defined in RDF/S (does not use OWL). In Beehive object model, there is a “parent” attribute defined for all entities (inherited from the Entity class). The cardinality of parent attribute in Entity is 0 or 1.

Siegfried commented that Nepomuk ontology uses the cardinality nrl:minCardinality and nrl:maxCardinality from NRL, which is a RDF/S extension. NRL addresses several limitations of current Semantic Web languages, especially with respect to modularization and customization.

Eric commented about the subtlety of modeling the “closed” or “open” collection of containers or parents of an item. In an integrated system, you can collect a complete or closed list of containers, each of which contains the item in question. For the WWW, the list of containers of an item needs to be an open collection since there is no way to ascertain that the collection represents a complete list. This is typically the case when sioc:has_container property is inferred from its inverse property sioc:container_of, i.e. one may not know all containers in all WWW sites that refer to the item.

Siegfried commented that most collections in Nepomuk desktop are closed lists. There is also a way to explicitly represent open lists in Nepomuk for things imported from Internet. Siegfried referred to http://www.semanticdesktop.org/ontologies/nrl/#2.3.3_Grouping_resources for ideas for collection and container in NRL and RDF/S.

Eric commented that for object-oriented model, we may introduce a construct to explicitly represent open collections (e.g. dynamic queries/searches) in addition to the regular closed collections. The container attribute in OO model may be represented as dynamic queries/searches to imply that the collection may not be complete, i.e. there may be other members that should be in the collection. For the desktop or integrated environments that assume closed world, we may introduce another attribute (perhaps called parent) that provides a complete list of parent containers.

The participants briefly reviewed some specialized containers, including forum, calendar, inbox, address book, and conference. Siegfried indicated that Nepomuk desktop uses the calendar ontology described in  http://www.semanticdesktop.org/ontologies/2007/04/02/ncal/ based on iCalendar. Nepomuk also uses email ontology described in http://www.semanticdesktop.org/ontologies/2007/03/22/nmo/. Nepomuk has a Thunderbird plug-in for semantic email, which uses the email ontology.

Eric commented that each of the specialized containers shall be reviewed one by one in subsequent TC meetings.

Siegfried requested a discussion thread to respond to Eric’s recommendation about ICOM TC defining a new, integrated, contiguous, and coherent model rather than importing the existing ontologies. Eric will open an issue and start the discussion thread for this issue.

4. Discussions of JPA and Jenabean prototype for ICOM

Eric indicated that it should be feasible to provide a prototype of ICOM through JPA and Jenabean interfaces on top of Beehive.

Siegfried explained that there are several implementations of Nepomuk for different platforms. The implementation for KDE is developed in C/C++. Nepomuk primarily used RDF as a medium between applications that may be developed in different languages.

The TC Meeting was adjourned.

Regards,

Eric

icom message