Olivier, all-- I was wondering if anybody knows of a mapping from IPTC Media Topics to Wikipedia categories?
http://cv.iptc.org/newscodes/mediatopic I found this question here, but no good answer: http://answers.semanticweb.com/questions/11457/anyone-published-mappings-between-dbpedia-and-subject-vocabularies-like-iptc-ddc-lcsh Cheers, Pablo On Fri, Jun 8, 2012 at 5:11 PM, Olivier Grisel <[email protected]>wrote: > 2012/6/8 valentina presutti <[email protected]>: > > Luca, I am forwarding this to stanbol mailing list, please next time use > this one. > > If you're not yet subscribed, please do it :) > > Good reflex to use the ML. Here is a copy of my reply: > > On 8 June 2012 12:34, Luca Cervone <[email protected]> wrote: > > Dear Oliver, > > I'm Luca Cervone from University of Bologna. > > I will take part to the IKS meeting next week in which I'll present our > > facebook game developed with the Prof. Valentina Presutti. > > This game uses the stanbol engines in order to create a resource for the > > sentiment analysis. > > We are trying to improve the game and for this reason we are evaluating > the > > possibility to use the topic classification engine. > > So we seen that the issue 197 (the rest API) is still open. Can you, > kindly, > > give us news about the status of the work? > > It mostly works but still lacks some web UI and documentation. > > > Is it possible to have a developing version of the API so we can install > it > > in our personal stanbol instance? > > Yes you can build and deploy the enhancer/engines/topic and > enhancer/topic-web bundles on an stable or full distribution of > stanbol. The you can use the webconsole to create a configuration for > a new classification engine and a matching training set (leave the > solr server parameters empty for a default configuration that should > work). > > Then you can have a look at the topic-web java source for the JAX-RS > endpoint to guess the HTTP API of the classifier model. > > You will need to use this HTTP API to import some possibly > hierarchical target concepts into the model using either individual > HTTP POST queries for each concept or a SKOS RDF taxonomy. > > Then use the API to upload many categorized text examples for each > registered concept. Then another call to train the model. > > There are some python script that do this for IPTC NewsML files or a > corpus of categorized articles extracted from DBpedia. > > > http://svn.apache.org/repos/asf/incubator/stanbol/trunk/enhancer/topic-web/tools/ > > Here is sample training corpus pre-extracted using the dbpediakit tools: > > > https://dl.dropbox.com/u/5743203/IKS/dbpedia/dbpediakit-output/dbpedia-taxonomy.tsv > > https://dl.dropbox.com/u/5743203/IKS/dbpedia/dbpediakit-output/dbpedia-examples.tsv.bz2 > > -- > Olivier > http://twitter.com/ogrisel - http://github.com/ogrisel >
