You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@stanbol.apache.org by Umutcan Şimşek <um...@mni.thm.de> on 2015/07/13 18:27:48 UTC
Entity Linking with Custom Vocabulary
Hi,
I'm trying to use EntityhubLinkingEngine with my custom vocabulary
Entityhub site, I cannot get any enhancement from partial matches. For
example my vocabulary has a person named Peter Smith. When I try to
enhance a sentence like "Mr. Smith works in this department." I don't
get any result from the entity linking engine. But dbpedia engines
sometimes return results from this kind of partial occurrences.
Is there a way to configure entityhub linking engine for this purpose?
One way to achieve my goal seems to be extracting given names and family
names as rdfs:label. I wonder if there is some other way.
Another question is, I'm using default solr index for my solryard. My
custom vocabulary is subject to change rather frequently. I read about
how to use genericrdf tool for this purpose but I think I need RDF dumps
for that. How can I create indexes?
Thanks
Umutcan
Re: Entity Linking with Custom Vocabulary
Posted by rh...@gmail.com.
Hi Umutcan,
On lunes, jul 13, 2015 at 6:28 p. m., Umutcan Şimşek <um...@mni.thm.de>, wrote:
Hi,
I'm trying to use EntityhubLinkingEngine with my custom vocabulary
Entityhub site, I cannot get any enhancement from partial matches. For
example my vocabulary has a person named Peter Smith. When I try to
enhance a sentence like "Mr. Smith works in this department." I don't
get any result from the entity linking engine. But dbpedia engines
sometimes return results from this kind of partial occurrences.
First, in that concrete sentence, you would to check if ’Smith’ has been recognized or not as a linkable token. You can try with other sentences to check if you get it linked. Anyway, probably you are going to need to trick the EntiyLinker configuration. Apart from the configuration that you can see in the OSGi console, there are several more parameters that you can adjust for your use case that are explained here:
https://stanbol.apache.org/docs/trunk/components/enhancer/engines/entitylinking#entity-linker-configuration
You can change these parameters directly in the engine configuration file
Is there a way to configure entityhub linking engine for this purpose?
One way to achieve my goal seems to be extracting given names and family
names as rdfs:label. I wonder if there is some other way.
Another question is, I'm using default solr index for my solryard. My
custom vocabulary is subject to change rather frequently. I read about
how to use genericrdf tool for this purpose but I think I need RDF dumps
for that. How can I create indexes?
If your vocabulary is going to change quite frequently, probably your best option is to use a ManagedSite (https://stanbol.apache.org/docs/trunk/components/entityhub/managedsite.html) instead of a ReferencedSite though the generic indexer tool. In that way, you can push your entities using the EntityHub REST API
Hope that helps. Cheers,
Rafa
Thanks
Umutcan