Dear Sebastian,
thank you very much for your detailed answer, and apologizes for
getting back only now, (I'm traveling: holidays!)
Sebastian Schaffert <
sebastian...@gmail.com> a écrit :
> Dear François,
>
> I'll try to answer to your comments inline...
[snip]
>> I installed the LMF (deploying the war in Tomcat 7.0.26), now following
>> the first steps and reading through the documentation. Cool! Some very
>> early basic remarks and questions after just a couple of hours of use:
>> - I first installed the 2.2.1, then upgraded to 2.2.2. When deploying the
>> 2.2.2 after un-deploying the 2.2.1 with tomcat manager, I got the cursor
>> spinning for a very long time… long enough for me to stop waiting. After a
>> computer restart, everything seems OK.
>>
>
> Upgrading currently does not really work. It might be necessary that you
> clean up the installation manually (in case of a webapp installation,
> remove the directory from tomcat/webapps, in case of a standalone
> installation, remove the whole installation directory).
it seems that things work if you undeploy, restart (tomcat only or the
machine?), re-deploy
[snip]
>> - Core Services / first steps: PUT replaces the whole metadata for the
>> resource. Is it possible to add triples to the description of a resource?
>>
>
> Currently only via a SPARQL Update.
ok
I have to look in more details to see what happens. Just tell me: what
is the answer of the query? Is it "A Game of Thrones"? (I have a blue
cover image for it)
>
>
>>
>> - Mike and Suzan's problem: returns all the 10 books. My guess is: the
>> solr search is an "OR" one, not an "AND" one. (If you remove "book" from
>> the "third book of "A Song of Ice and Fire" search string, you get 4
>> results only, which is better)
>>
>
> Yes, the SOLR search is an OR query. In most cases this is also the desired
> behaviour (e.g. in Google), because relevance is determined by ranking, but
> we'll look into it to improve the demo.
Ah, OK. it looks like I'm not used enough to lists of results ordered
by ranking when it is about sw resources (too used to standard sparql
queries)! The right answer is indeed the first one in the list. Well
done
[snip]
>> What I'd like to do now: I have a LD compliant app running on my machine
>> and that I use very frequently: a tagging tool, where tags are SKOS
>> concepts (cf.
www.semanlink.net). What is the easiest way to connect it
>> and LMF? How could I easily say that any of my tags (of the form
>>
http://127.0.0.1:8080/semanlink/tag/foo) correspond to an LMF resource,
>> eg.
http://127.0.0.1:8080/LMF/resource/tag/foo? (In the future, I could
>> considerer the replacing of my aging application by something build using
>> LMF / VIE / IKS - but for the moment, I'd just want to test things, have
>> some real content to play with inside LMF and get a clearer idea about what
>> and how I could do)
>>
>
> The LMF actually goes beyond a Linked Data Server, so you could even get
> representations of resources like
http://127.0.0.1:8080/semanlink/tag/foo in
> the LMF.
I had done something with similar results (I think), creating a
resource
http://127.0.0.1:8080/LMF/resource/tag/foo, and then
describing it using an owl:sameAs
I think you meant:
http://127.0.0.1:8080/LMF/resource?uri=…
it returns a kind of copy of the html returned when dereferencing
http://127.0.0.1:8080/semanlink/tag/foo (and requesting html): the LMF
seems to get the resource with an accept header set to receive html -
this is not bad, but would it be possible to decide between this html
returned by the original resource and the one that the LMF would
compute from the triples returned when getting the original resource
with an accept header set to return RDF (the 3-column view) ? Related
questions:
- you make this "copy" of the original html
- Why not just redirecting to the original resource?
- this "copy" may require some complicated transformation (eg. if
there are relative links in the html): I'm not sure you handle that
correctly in every case. For instance, I have in my HTML:
<link rel="foaf:primaryTopic" href="/semanlink/tag/foo" />
and it remains unchanged in the "copy" html (which is incorrect)
To make a "clean" copy, you have to parse the html, and this is not an
easy task. I had used once javax.swing.text.html, but it has several
bugs. What do you use?
- in the LMF "standard" description of a resource (the 3-columns
table), when clicking a resource, you sometimes gets the same
3-columns view, and sometimes the HTML returned by the resource. Do
you always return the HTML view of the original resource when there is
one, or does it depend on something else?
When dereferencing
http://127.0.0.1:8080/semanlink/tag/foo, you get
triples where this resource is subject, but also triples where it is
object (and other statements also). When getting the corresponding
http://127.0.0.1:8080/LMF/resource?uri=…, you only get the triples
where it is subject. Can I still have access to the other ones?
>
> 3. repeat the same POST for other tags (e.g. using a script); unfortunately
> Linked Data does not provide any means to list resources, so it needs to be
> done in this way for now
>
> 4. use it e.g. for configuring a semantic search over tags; you will have
> to turn the Semantic Search configuration option "solr.local_only" to false
> (by default the LMF only considers resources for search where it thinks it
> is the "owner" and then you can define a search core as usual).
>
> Hope it works!
I have not tried it yet. Clearly I'll need to index my resources also
using statements where they appear as objects. But, as I said, I'm
traveling, and it is too hot to have my computer turned on for too
long! (I once got a melted battery and don't want to repeat the
experience). I didn't want to wait more before sending an answer to
your message. I'll tell you later.
>
> Sebastian
>
Thanks again,
fps