[BangPypers] [Inpycon] Weekend meeting
Anand Balachandran Pillai
abpillai at gmail.com
Sat Oct 10 18:50:08 CEST 2009
On Fri, Oct 9, 2009 at 11:23 PM, abstract things
<abstractthings at gmail.com>wrote:
> Hello All,
> My name is Sundar and I am Senior Python programming language user [?] (I mean 3+ years). I like Python programming. Well, I haven't registered for PyCon early enough but I attended Mayavi, Algorithms and Semantic Web talk. Wanted to attend many other interesting talks as well but time didn't permit me. So following some of the online material now.
> I also have very much interest in Semantic Web but I some how don't agree with the Anand's post. No offence Anand. I am following Semantic Web since 2005 onwards (soon after the SPARQL draft was out) and have been active member in online Semantic Web groups. Open Calais is a Web service which stores data/information with Semantics. When we access Open Calais store by its APIs, what we get in return is JSON format (may be others as well, I haven't explored much its API) results with entity and value etc. But that is something like taking data out from your database after sending a query. If somebody follows WordNet and NLP can figure it out with document classification algorithms that the particular text you are sending to Open Calais has what semantics associated (associated Nouns, Verbs e.g. Person, place, actions, class of the text i.e. sports, politics, history etc).
> Arranging this information on our webpage or blog is not the Semantic Web. I would call it as fetching data from a Semantic Data store and arranging it on our webpages so that Search engines can find it in better manner or we can have quality or enriched information on our webpages (one can use RDFa). Semantic Web is what happening inside Open Calais along with other Linguistics based features for extracting meaning of the data we send or ask through API. And no way anybody can call this as ONTOLOGY. Ontologies are like domain modelling, a hard thing). Ontologies are inside thing of Open Calais. What we get out from their data store is not an Ontology. Its simply the data with semantics we are getting (just the relationships without actually knowing the domain model built inside). We no way have any knowledge of how the data inside Open Calais is stored and what their ontology for a particular domain is (Something similar to getting results out from a database without knowing what the internal schema is).
> Semantic Web basic concept has RDF (which is in turn an XML based standard). What that guy in PyCon showed in his presentation was a direction to think about Semantic Web if you yourself is building some Semantic Web based system (eg. if your company wants to launch some online system with Semantic Web features then you can have your domain vocabulary defined and accessible using a URL, also how the RDF statements can be made, what are the best practices for making such statements etc). His talk was more of focused on large Enterprise systems and integrations compared to using Semantic Web for Web page based data annotation. (Micro formats is not the part of W3C based Semantic Web protocol stack).
> Based on my limited knowledge, RDFa and GRDDL are the two standards from W3C that can be used for Web content annotation and the RDFLib (that guy used this library in his practical Python examples if you remember) can generate RDFa documents from the triples you have in your datastore (see his last 3 or 4 slides). RDF based XML documents can be generated from a Web application or an enterprise application for semantic based data exchange.
> Well, my thoughts here are not to argue on any thing but to give a picture of Semantic Web from my acquired knowledge in last 4or so years. (and also the perspective of that guy as I observed).
> I guess it would have been more interesting if that guy would have given few examples of using Description Logic to describe data relationship and how OWL fits into that model and also about Stanford's Semantic Web Protege editor or some other editors for designing Ontologies. I guess it requires more than 45 min.
> Nevertheless, his talk was satisfying for me, may be because I already had background knowledge of this subject.
> On Fri, Oct 9, 2009 at 4:48 PM, Arvind Jamuna Dixit <ardsrk at gmail.com <http://mail.python.org/mailman/listinfo/inpycon>>wrote:
> >* Semantic web and python. Pretty cool stuff. I am interested.*
> Cool. Here is a brief of what I want to present.
> Anyone who attended the talk on Semantic web in Pycon would have
> understood something about the "theorotical" semantic web as
> visualized by W3C.
> This is however a top-down approach which requires annotating
> web-pages with metadata such as RDF, microformats etc.
> However a more practical approach is extracting
> semantic content namely ontologies and relationships from existing
> web content. Such services are already becoming available.
> I will be demonstrating such a web-service namely "OpenCalais"
> by Reuters and how to use it in Python to develop some applications
> using the rich semantic data returned by the service.
> The whole session will be interactive and about showing running
> Python code, and will last from 45min - 1 hour.
I am sorry, but I decided not to present this talk tomorrow. I had thought
of presenting OpenCalais as an example of top down semantic web but
with some prejudices already about the topic, I think I will be better
off with presenting something more than just that.
I am working on an idea which I have to use the service to build up
a semantic web application which uses OpenCalais to extract
data and present it via other web 3.0 services. However it won't
be ready by tomorrow and I dont have the time to hack something
up quick. Next week-end I will have something to present on this
and I think it will be better to present everything together than
piece by piece.
So this talk is canceled tomorrow. Let us have the PyCon
-------------- next part --------------
An HTML attachment was scrubbed...
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 96 bytes
Desc: not available
More information about the BangPypers