Hi Rafa, Thank you for your very extensive feedback.
A SkosJS integration with Stanbol is the most logical and that's what I tried already some time ago. Unfortunately, SkosJS creates entities using SPARQL update queries, which are not supported by Stanbol. This integration seems difficult to me at the moment. That's why I chose Marmotta to integrate with SkosJS. I went for your suggestion to use a ManagedSite based on a SolrYard. As a quick solution, I was able to create a little sync script in bash that exports the content from Marmotta into an RDF file and then imports it through the Stanbol REST API. This works great, but probably only for small datasets (which is our case right now). This will enable us to make progress, but we'll probably need to create a Marmotta module to push changes to Stanbol. Thank you for you help! -- Mano On Fri, Aug 7, 2015 at 1:27 PM, Rafa Haro <rharoapa...@gmail.com> wrote: > Hi Mano, > > > > > I understand what you mean now. You are making a wrong assumption about > Stanbol. The entityhub Referenced Site local cache doesn’t work that way. > It doesn’t automatically synchronize with the remote site. As any other > cache, as far as I know, it works by storing locally those remote entities > that you have retrieved through dereferencing at any time. That means that > when you configure it for the first time, Stanbol is not going to retrieve > all the entities by using a kind of SELECT ?s ?p ?o SPARQL query. > > > > > The full synchronization with a TripleStore is, in my opinion, an > extremely interesting use case, because it is very natural, as you are > doing right now, to store RDF data in a triple store and have it > immediately available for enhancing. The thing is, the SolrYard is the only > Yard that really works for Enhancing. So, there should be a way to > synchronize a TripleStore backend with an Stanbol SolrYard, but this is not > easy to architecture from the Stanbol point of view and also would imply to > couple with a concrete TripleStore where you need to include the pushing > module to Stanbol. > > > > > So far, the best option would be to use a ManagedSite with a SolrYard > instead of a ReferencedSite and use the EntityHub REST API for pushing > entities. You are using SKOSjs, maybe you can extend it for communicating > directly with the ManagedSite and I suppose that wouldn’t be > straightforward also because any editing action at SKOSjs that finally rely > on concrete SPARQL queries must have a correspondent REST call to the > EntityHub > > > > > Cheers, > > Rafa > > On Fri, Aug 7, 2015 at 11:23 AM, Mano Swerts <mano.swe...@aca-it.be> > wrote: > > > Hi Rafa, > > Thank you for the reply! > > I have no issue with the fact that a local index is required, but the > local > > index should update automatically (or check for changes in a cron) to > make > > sure that it is up to date. It is not useful if the client needs to > contact > > us every time they make a change so that we can update the indexes. > > I also saw a module in the Stanbol source code called jennatdb. Is this > > maybe an option? I have no problem with switching to Jena instead of > > Marmotta. > > Hopefully somebody else in this mailing list can help me. > > Kind regards. > > Mano Swerts > > On Fri, Aug 7, 2015 at 11:07 AM, Rafa Haro <rharoapa...@gmail.com> > wrote: > >> Hi Mano, > >> > >> > >> > >> > >> I have never tried to configure a ReferencedSite without a local index > >> (i.e. using only the remote dataset) so I couldn’t help you right now > too > >> much. According to the documentation, apparently the remote site is used > >> only for dereferencing but can’t be used alone for enhancing (for > searching > >> for entities using SPARQL). > >> > >> > >> > >> > >> Let’s see if someone else can shed light on this. Sorry :-( > >> > >> On Thu, Aug 6, 2015 at 5:25 PM, Mano Swerts <mano.swe...@aca-it.be> > wrote: > >> > >> > Hi all, > >> > I want to use Stanbol to enhance content, but I do not succeed in > setting > >> > it up. We use the following tools: > >> > - *SkosJS* (for a non-technical user to manager a taxonomy) > >> > - *Apache Marmotta* (contains the data. It is linked to SkosJS. It > was > >> > not possible to link SkosJS with Stanbol) > >> > - *Stanbol* (use the data in Apache Marmotta to enhance content) > >> > I registered Marmotta as a Referenced Site through SPARQL. My entities > >> are > >> > available through this Referenced Site. > >> > I created an Enhancer Engine coupled to the Referenced Site, which is > >> added > >> > to the default chain. It is detected and used by Stanbol, but I get no > >> > results when enhancing content. > >> > I looked at the existing DBPedia setup and noticed that it uses a Solr > >> Yard > >> > and Cache. I think this might be the clue, but when I link my > Reference > >> > Site to my newly created Cache it is suddenly not available anymore. > >> > Therefore the enhancers won't work. > >> > I simply created the Solr Yard and Entityhub Cache through the Apache > >> Felix > >> > Web Console. > >> > I noticed that there is an folder called "indexes" which contains > indexes > >> > from DBPedia, but none for my own Yard. I also read something about > >> > generating indexes manually, but in our use case this does not seem > >> > feasible. When users add new data using SkosJS, it should be > automacially > >> > made available to the Enhancer. > >> > Can somebody help me with this? > >> > Many thanks. > >> > Mano Swerts > >> >
