[PragmaticWeb] Faceted Search

Sun Dec 6 14:59:22 CET 2015

Sebastian,

I've looked at the code on Git. I think we've gone through this design
phase as well: a custom object model.

If you constrain yourself only to the standard RDF model and
manipulate it using SPARQL (or an RDF API such as Jena or Sesame),
then you can map the rest of data models into it using vocabularies.

These class names look like they're taken from the Jena's codebase,
just oversimplified and with much poorer functionality:
http://sourceforge.net/p/cognescent/repository/ci/master/tree/src/net/java/cognescent/sbi/model/onto/

In other words, domain model at the object layer is not necessary at
all if you use RDF. And custom APIs are not necessary as you can reuse
generic open-source libraries. Constraint fosters creativity :)

They say Linked Data should not be treated as a hammer for which every
problem appears as a nail. But it's the first hammer which includes a
manual on how to turn problems into nails -- and them you hammer them
down with ease :)

Martynas
graphityhq.com

On Fri, Dec 4, 2015 at 5:55 PM, Sebastian Samaruga <cognescent at gmail.com> wrote:
> Martynas,
>
> The sources are in the Git repository of the Sourceforge project. They are
> in the form of an importable Eclipse project, it is not yet a Maven project.
> Once you import the project in Eclipse from the repository (from the
> repository browser of Eclipse put "Import as a project in the workspace")
> you must set some library paths to make it work in your environment.
>
> I'm not using the relational model, it is just an analogy.
>
> Sebastian.
>
>
>
>
> On Fri, Dec 4, 2015 at 5:30 AM, Martynas Jusevičius <martynas at graphity.org>
> wrote:
>>
>> Sebastian,
>>
>> as I mentioned before, a .war file is a poor channel of distribution these
>> days. Please show us your source code and/or deployed application.
>>
>> Are you using the relational model or your own metamodel for data
>> integration? This paper explains the advantage of RDF and triples as the
>> canonical model:
>>
>> http://www.topquadrant.com/docs/whitepapers/information-enlightenment-2.0-final.pdf
>>
>> By putting RDF at the center, you could reuse the RESTful Linked Data
>> processing from Graphity and focus on BI which can be implemented with
>> SPARQL.
>>
>>
>> Martynas
>> graphityhq.com
>>
>> On Wed, 2 Dec 2015 at 20:20, Sebastian Samaruga <cognescent at gmail.com>
>> wrote:
>>>
>>> Martynas,
>>>
>>> Happy you replied my message. I've been watching Graphity and it is
>>> awesome. But, although our scopes seems to be similar I think there are huge
>>> differences at implementation level and at presentation level.
>>>
>>> What I would like to have built is a 'semantic' browser focused in
>>> analysis and mining for business intelligence and for building indexes and
>>> process flows definitions. I use forms for that and I also use XSL but for
>>> data coming from my metamodel, serialized from JAXB beans. My metamodel is a
>>> completely normalized scheme in which 'tables', 'columns' and 'values' are
>>> interchangeable entities aggregated into 'rows' (Mappings). It comes from a
>>> variety of sources, including but not just RDF. The metamodel has a notion
>>> of triples regarding the application state and I use them when triple
>>> exchange is needed.
>>>
>>> I'm aiming to be able, for example, to build queries with analysis
>>> metadata from the URIs given to a REST service. Being this service a HATEOAS
>>> service, a client could browse through the use cases / contexts only with
>>> metadata. Loaders from multiple sources (JDBC, RDF, XML, JSON, CSV, SPARQL,
>>> etc) exist to populate the mappings layer. The browser UI aggregates
>>> 'tables' (metaclasses), once a metaclass is selected it shows its 'columns'
>>> (roles) and selecting a 'column' shows the column values. If I select one
>>> 'row' while navigating (instances) the values shown will be only the ones of
>>> that instance (Mapping).
>>>
>>> You can check it in Sourceforge
>>> http://sourceforge.net/projects/cognescent/
>>>
>>> The WAR file downloadable there shows examples with a sample load data
>>> form. Modify datasources.xml in the src root to test other datasources. The
>>> XSL templates (Resumen button in the browser) only outputs the raw XML. Once
>>> the stylesheets are developed they'll bring a cleaner knowledge view.
>>>
>>> A sample URI to check in the REST interface would be like (dummy):
>>>
>>> http://localhost:8080/CognescentSBI/resources/data/metaId:instId[predAxisId]/instId:instId[predAxisId]/instId:instId[predAxisId]
>>>
>>> To test the REST service go to:
>>> [your_server:your:port]/CognescentSBI/resources/sbi/data
>>>
>>> The predicate axis IDs (predAxisId) will be for the implementation of the
>>> Analyzer component so it can aggregate instances due to this predicates
>>> values. The URI is parsed to build the faceted queries.
>>>
>>> Best,
>>> Sebastian.
>>>
>>>
>>> On Wed, Nov 18, 2015 at 1:29 PM, Martynas Jusevičius
>>> <martynas at graphity.org> wrote:
>>>>
>>>> Sebastian,
>>>>
>>>> there was little traffic on this list so far, yet your message went
>>>> unnoticed. Sorry for that!
>>>>
>>>> You can view the webpage for your data as an (X)HTML representation of
>>>> RDF. It can be completely data-driven and simply a function that
>>>> transforms the RDF response into an (X)THML response. You could do it
>>>> in a number of ways: using an imperative language such as Java, or
>>>> various template engines.
>>>>
>>>> From a declarative perspective, XSLT (2.0) is probably the best tool
>>>> for this. The RDF/XML syntax is admittedly not straightforward, but
>>>> XSLT is a functional and Turing-complete language that can handle it
>>>> effectively and transform it to (X)HTML or other formats.
>>>> Graphity Client includes a number of predefined generic XSLT
>>>> stylesheets that can be imported, extended, customized etc:
>>>>
>>>> https://github.com/Graphity/graphity-client/tree/master/src/main/webapp/static/org/graphity/client/xsl
>>>>
>>>> The facets part is more complicated. If you are using SPARQL to
>>>> retrieve RDF results, you can modify the query on the fly, to narrow
>>>> down matches based on the facet selections. That is what we did in
>>>> this project: http://dedanskeaviser.dk/newspapers
>>>> If not, then it depends on the API you are using to access RDF.
>>>>
>>>> I am not familiar with DCI or its concept of actors/roles/contexts.
>>>> Maybe it would be easier if you explained your goals. If you are
>>>> looking for a functional implementation of data-driven applications
>>>> that uses REST and HATEOS and builds interactive forms and other
>>>> navigation items, it is exactly what we provide:
>>>> http://graphityhq.com/technology
>>>>
>>>> We would be happy to this topic discuss further.
>>>>
>>>>
>>>> Martynas
>>>> graphityhq.com
>>>>
>>>> On Mon, Oct 12, 2015 at 10:26 PM, Sebastian Samaruga
>>>> <cognescent at gmail.com> wrote:
>>>> > Sorry if this is not the right list. I've came here from a reply in a
>>>> > previous post.
>>>> >
>>>> > I don't know if this is something new, it just seems it was useful for
>>>> > the
>>>> > kind of interface I was looking for. I was just wondering what would
>>>> > be the
>>>> > best way to facilitate browsing and search in the application demo I'm
>>>> > building that renders RDF / Semantic Web contents. I've figured out
>>>> > there
>>>> > must be a tree hierarchy of categories, roles and instances of data to
>>>> > which
>>>> > adhere the incoming data parsing so having a common denominator for
>>>> > different input structures.
>>>> >
>>>> > Having this structures, browsing through the tree of data, an item
>>>> > (leave or
>>>> > node) could be 'picked up' as a facet. For example, if the scenario is
>>>> > "Car
>>>> > Rental" as a category, "Car Model" and "Rental City" as roles and many
>>>> > models of cars and many cities as instances, what if I could pick a
>>>> > car
>>>> > model, a city or both and press "Aggregate" and this resulting in root
>>>> > categories for each specific car rental ("Car Rental 1", "Car Rental
>>>> > 2",
>>>> > etc) with its roles populated with the corresponding criteria values
>>>> > (the
>>>> > city corresponding to "Car Rental 1" given its car, etc).
>>>> >
>>>> > Maybe this sounds dumb. But the question is: how difficult would be to
>>>> > build
>>>> > such a filter criteria using only RDF datasources. RDF statement
>>>> > resources
>>>> > are not individualized by their occurrences. An RDF resource is the
>>>> > same
>>>> > regardless which statements it occurs. And, although I've found a way
>>>> > to
>>>> > individualize occurrences of, for example, Car Rental(s), I can't find
>>>> > the
>>>> > way yet to correlate this instances with the instances of their roles.
>>>> >
>>>> > Also, I'm restricting my mappings (ontology processing output) to
>>>> > three
>>>> > levels depth, which seems arbitrary. I could not restrict the graph to
>>>> > any
>>>> > depth. But I'll keep trying a while with this arrangements. It seems
>>>> > attractive the correlation of categories, roles and instances with
>>>> > some of
>>>> > the concepts in the DCI programming model
>>>> > (https://en.wikipedia.org/wiki/Data,_context_and_interaction) which
>>>> > could
>>>> > allow for a model driven focused approach of building the client
>>>> > application, again with the ontology 'common factors' concept in mind.
>>>> >
>>>> > And the concept of actors, roles and contexts seems as an ideal case
>>>> > for a
>>>> > functional language implementation of what could be a runtime
>>>> > environment
>>>> > for data driven applications (think of a REST-HATEOAS client that
>>>> > interactively builds forms and other navigation items simply by
>>>> > content
>>>> > negotiation with a functional data-state-based endpoint).
>>>> >
>>>> > Source, examples and the demo web application are available in the
>>>> > project
>>>> > web page:
>>>> >
>>>> > http://cognescent.blogspot.com
>>>> > http://sourceforge.net/projects/cognescent/
>>>> >
>>>> > Regards,
>>>> > Sebastian
>>>
>>>
>