Hi Rupert,

Thanks for your feedback. My response below.

On Thu, Jul 5, 2012 at 6:40 AM, Rupert Westenthaler <
[email protected]> wrote:

> Hi Ivo,
>
> > So here are some questions:
> >
> > 1. What could be added to EntityHub from the knowledge listed above and
> > what is the best (stable enough) way to add entities to Entity Hub: is it
> > via REST or manually as described here:
> > http://incubator.apache.org/stanbol/docs/trunk/customvocabulary.html (of
> > course REST would be more interesting, but if it is not ready yet a
> manual
> > approach would be good too)
>
> I have already started to implement STANBOL-673 [1] that will bring a
> new type of "Site" to the Entityhub that can be fully managed by the
> RESTful API. This is exactly tailored to use cases as described here.
> In a 2nd iteration of this I do also plan to make all the
> functionality of the Entityhub Indexing tool (as described in [2])
> available for ManagedSites.
>

Ok, as we are evaluating the project right now we need something that works
now :)


>
> In the meantime you could use the RESTful services of the Entityhub
> (http://{stanbol-server}/entityhub/entity)
>


ok, I found this links http://dev.iks-project.eu:8081/entityhub as a
reference, is there any other docs about this?


>
> > 2. Would we need to tackle Onthology Manager in any way to organize the
> > entities or this can be skipped? (this is the most vague thing I have
> > encountered)
>
> It really depends what you want to do. If you just want to use the
> Stanbol Enhancer (and the Entityhub for managing the Entities to
> extract from parsed Content) than you will not need to use the
> Ontology Manager, Reasonings and Rules component.
>

Ok, that simplifies things as we need to tackle one module less, we don't
have much time to do it anyway.


>
> > 3. How to connect enhancer to use our entities? Is the Keywordlinking
> > Engine way to go:
> >
> http://incubator.apache.org/stanbol/docs/trunk/enhancer/engines/keywordlinkingengine.html
> >  ?
>
> I would expect so. The use cases [2] and [3] should provide all
> necessary information to help you with that decision. If not feedback
> on how to improve those is very welcome.
>

Ok.


>
>
> > 4. Would it be faster to use the same entity types as dbpedia (Person,
> > Company,...) or introducing new ones should be straightforward?
>
> Personally I would suggest to use
>
> * schema.org [4] for entity specific information - because you might be
> able to use the schema.org mapping also for SEO
> * SKOS [5] for describing the hierarchy and relations between concepts
>

Ok. so we would be adding SKOS in RDF format to entityhub, and
schema.orgoptionally for specific info.


>
> Some additional comments/suggestions:
>
> Note: I copied lines multiple times and also re-ordered them to better
> fit to my recommendations.
>
> > - objects are instantiated from a class (there is a collection of default
> > classes but it is possible to add more)
> > - class has collection of attributes
> > - objects can have relations with other objects
> > object1 is_a_class_of class1 (e.g. Person, Blog post, Folder, ... )
>
> Content Objects that use classes representing Entities - Person in the
> above list) would be very useful for adding to the Entityhub (or a
> ManagedSite as soon as they are available). As mentioned above I would
> try to map those classes and attributes to schema.org
>
> > - objects can be tagged with tags
> > - tags are organized separately and can be hierarchical too
> > object1 is_tagged with tag1
> > tag1 is_a_parent_of tag2
>
> Having a tag hierarchy AND tagged documents is ideal for
> training/using the TopicEngine. See [6] for how to train/use this
> engine. So this data can be used to provide an auto-tagging feature
> for your CMS.
>

ok, that is interesting, we will check this out.
Btw, is this Topic Engine using solr/cluster feature internally?

To wrap it up, we would need to:
- do the stanbol installation and configuration
- define what entities we are going to push to stanbol
- make a SKOS exporter for CMS
- make a servise that syncs the SKOS info to entitiyhub REST interface
- configure the keyword linking engine to use our entities
- use the enhancer rest api to get the enhancements via keyword linking
engine
- optionally explore the Topicengine

Are we missing some step or we could define a project with this and make a
proof of concept?

Cheers






>
> > - objects are stored hierarchically as nodes
> > object1 is_a_parent_of object2
> > object1 is_related_to object3
> > object1 is_a_class_of class1 (e.g. Person, Blog post, Folder, ... )
>
> best
> Rupert
>
> [1] https://issues.apache.org/jira/browse/STANBOL-673
> [2] http://incubator.apache.org/stanbol/docs/trunk/customvocabulary.html
> [3] http://incubator.apache.org/stanbol/docs/trunk/multilingual.html
> [4] http://schema.org/docs/full.html
> [5] http://www.w3.org/2009/08/skos-reference/skos.html
> [6]
> http://dl.dropbox.com/u/5743203/IKS/ReviewMeeting2012/Topic-Classification.pdf
>
>
> On Wed, Jul 4, 2012 at 10:54 AM, Ivo Lukač <[email protected]> wrote:
> > Dear Stanbol Community,
> >
> > I work at Netgen, a small web agency mostly using eZ Publish CMS and we
> are
> > also an eZ Publish  business partners for a long time. Together with
> > another independent eZ consultant Paul Borgermans we are interested to
> add
> > some semantic possibilities to it so we were in Salzburg last month to
> try
> > to figure out what exactly could we do. My colleague Petar followed up
> the
> > discussion with Mr. Suat Gonul after the event.
> >
> > After seeing some solutions already been made by early adapters and
> knowing
> > the eZ Publish CMS architecture really well we have a potential idea on
> > what we could do as a proof of concept. But we need your help to evaluate
> > the idea and give some feedback.
> >
> > The idea is simple: to map the eZ taxonomy to Stanbol and use the
> > enhancements to help the editor annotate the content.
> >
>
> > Of course, we could add more specific knowledge (depends on the specific
> > project) but we would like to keep it as general as can be. What we could
> > do easily is to generate an RDF with the data.
> >
> > So our main goal would be to push part of this knowledge to Stanbol and
> use
> > it to enhance newly created objects not using dbpedia but rather internal
> > data.
> >
> >
>
> >
>
> >
> >
> > 4. Would it be faster to use the same entity types as dbpedia (Person,
> > Company,...) or introducing new ones should be straightforward?
> >
> > Any kind of feedback would be welcome and would be helpful in our
> > evaluation....
> >
> > Best regards
> >
> >
> > --
> > Ivo Lukač
> >
> > Netgen d.o.o. - A.M.Tripala 3/I, 10000 Zagreb, Croatia
> > web: http://www.netgen.hr, tel: +385 (0)1 3879722, mob: +385 (0)91
> 5251566
> > ---------------------------------------------------------
> > everyday tweets: http://twitter.com/ilukac
> > company blog: http://www.netgen.hr/eng/blog
> > professional profile: http://www.linkedin.com/in/ivolukac
> > personal blog: http://ilukac.com/
> > member of the CISEx board: http://www.exportboomers.com/
> > presenting: http://ezsummercamp.com
>
>
>
> --
> | Rupert Westenthaler             [email protected]
> | Bodenlehenstraße 11                             ++43-699-11108907
> | A-5500 Bischofshofen
>



-- 
Ivo Lukač

Netgen d.o.o. - A.M.Tripala 3/I, 10000 Zagreb, Croatia
web: http://www.netgen.hr, tel: +385 (0)1 3879722, mob: +385 (0)91 5251566
---------------------------------------------------------
everyday tweets: http://twitter.com/ilukac
company blog: http://www.netgen.hr/eng/blog
professional profile: http://www.linkedin.com/in/ivolukac
personal blog: http://ilukac.com/
member of the CISEx board: http://www.exportboomers.com/
presenting: http://ezsummercamp.com

Reply via email to