Hello Jie Gao,
Can you send an example pipeline (including the documents in it) that
demonstrates these cases.It will be very helpful for understanding the
context.
Thank you,
Philip Alexiev
Software Engineer, KIM team
On 08/31/2012 01:16 PM, JIE GAO wrote:
Hi, ontotext team:
I am currently evaluating KIM 3.6-SNAPSHOT.
I've found that two KIM customized GATE PRs have strange behaviour for
me. The one is KIM OrthoMatcher and another is Instance Generator.
The scenario is that i customized many JAPE grammars based on KIM
default Jape grammar PR. There is a typical rule which is used to
extract person entity with title appearing in text. The default rule
in orginal KIM grammar is called "PersonTitle". The same as default
grammar definition, i set the "originalName" as person name, while
annotating the person entity combined the name with title. The
orthmatcher perform well for this practice. However, instance
generator failed to generate instance URI for the entity. Then, i
changed the grammar to set "originalName" to full name (e.g., "Miss
Putran" rather than "Putran"). This change makes the KIM OrthoMatcher
not working. In other words, the entity "Miss Putran" cannot be
matched with the entity labeled "Santosh //Putran", whereas, this time
"Miss Putran" can be generated with an URI by instance generator.
Based on my analysis from KIM sourcecode, KIM orthomatcher
will always use "originalName" to retrive the entity label which is
hardcoded and not changable. I have found the "stripPersonTitle"
function, whereas havn't found any usage in the sourcecode so far. The
experiment result turns out that the KIM Orthomatcher failed to strip
Person title from "originalName" , which cause the orthomatcher
failing to remove person title before matching.
Meanwhile, i have found that the instance generator will
firstly to check the duplication in knowledge base if there is matched
entity, it will be directly specified with the same URI . If there is
no same entity in the knowledge base, the entity will be processed by
"IEMetadataAppender" and set the context to be "http://newEntity=true"
if no specific one given. However, my test turns out that if the
original name is not the same as full name of specific extracted
entity (i.e., original name is "Putran" rather than "Miss Putran" and
"Putran" is not identified as a existing entity in KB). The instance
generator will not genrate URI for such kind of entity.
/Look forwards to your kindly help.
///
*Thanks & Regards*
JIE GAO
_______________________________________________
Kim-discussion mailing list
Kim-discussion@ontotext.com
http://ontomail.semdata.org/cgi-bin/mailman/listinfo/kim-discussion
_______________________________________________
Kim-discussion mailing list
Kim-discussion@ontotext.com
http://ontomail.semdata.org/cgi-bin/mailman/listinfo/kim-discussion