Hi
On Tue, May 14, 2013 at 12:51 PM, Joseph M'Bimbi-Bene <[email protected]> wrote: > Yesterday, Mr. Hernandez published a french model of OpenNLP, > https://sites.google.com/site/nicolashernandez/resources/opennlp > and i tried to integrate it with Stanbol to help with an entityLinking task. Thanks for the pointer! > > Here is what i get with talismane on an example token: > "ProcessingState > 0: Token: [0, 5] mario (pos:[Value [pos: > NC(olia:CommonNoun|olia:Noun)].prob=0.22461293861915013]) chunk: 'none'" > > And now with the french model of OpenNLP: > > OpenNlpPosTaggingEngine No POS TagSet registered for Language 'fr'. Will > build an adhoc set based on encountered Tags! > [...] The Stanbol NLP processing module requires to have a mapping of the String POS tags as used by the Tagset to the Olia Ontology concepts as sued by Apache Stanbol. ATM there is no way to configure such mappings. They are "hard coded" in the OpenNLP POS tagging Engine (see [1]). Making those configureable would require to write an parser/serializer for those configurations. There is also an other possibility. It is possible to configure String POS tags that are considered as linkable and matchable for the EntityLinkingEngine (see [2] and search for "Token level Parameters:"). The 'pos' parameter could be used to define that the string tag 'NPP' represents a linkable category of words in french The according line for the "Processed Languages" (enhancer.engines.linking.processedLanguages) would be fr;pos=NPP best Rupert [1] http://svn.apache.org/repos/asf/stanbol/trunk/enhancement-engines/opennlp/opennlp-pos/src/main/java/org/apache/stanbol/enhancer/engines/opennlp/pos/model/PosTagSetRegistry.java [2] http://stanbol.apache.org/docs/trunk/components/enhancer/engines/entitylinking > OpenNlpPosTaggingEngine > Token: [0, 5] Mario > org.apache.stanbol.enhancer.engines.opennlp.pos.services.OpenNlpPosTaggingEngine > - Value [pos: NPP([])].prob=0.7535221968807173 > > And then the linking just fails, every token has > "linkable=false(linkabkePos=false)| matchable=false(matchablePos=null)" > > What is the next step so i can make it work ? > <https://sites.google.com/site/nicolashernandez/resources/opennlp> -- | Rupert Westenthaler [email protected] | Bodenlehenstraße 11 ++43-699-11108907 | A-5500 Bischofshofen
