Dear Sonia,
As far as I can understand, Stanbol core mission is to provide an
integration framework, and not necessarily to provide state-of-art NLP
technologies. The problem you mention will pop-up probably in any
application context which uses an out-of-the-box model, be it OpenNLP or
whatever else. We are working in integrating also Linguagrid services in
Stanbol: this will provide the user with more options, probably allowing
her the implementation of a voting mechanism, but again, even linguagrid
WS are trained on specific corpora (mostly DBpedia, not surprisingly:-)
) and are consequently equally biased.
Cheers,
Luca
On 20/03/2012 09:30, Fabian Christ wrote:
Hi Sonia,
we have seen this reported by others. Maybe this thread is also
relevant for you:
http://mail-archives.apache.org/mod_mbox/incubator-stanbol-dev/201203.mbox/%3CFBEE806D-5578-4E75-9A15-5B1868103DBB%40open.ac.uk%3E
Best,
- Fabian
Am 20. März 2012 09:24 schrieb Sonia Gomez<[email protected]>:
HiI extract entity from text, but the relevance of extraction isn't very good, in my texte I have
some technical informations like "push On/Off" or "Vertical"Stanbol extract
this entity :People :
Referenced entitiesGeorge V of the United KingdomCharles V, Holy Roman
EmperorPhilip V of SpainMentionsV
PharaohReferenced entitiesPharaohJesusAl MichaelsMentionsOn
any idea to resolve this mistake. Thanks :)
