The IBUKI Named Entity Hierarchy

IBUKI uses IBML as the basis if its Named Entity Hierarchy.  IBML provides the
following features.

   IBML types sometimes behave like schemas
   Hierarchical organization with multiple inheritance
     multiple inheritance means that a thing may have 
       more than in type and they do not need to be comparable
     eg ass    names a kind of animal
        ass    names a body part 
     multiple inheritance means that the types
       a 3 toed 2 legged animal
       a 2 legged 3 toed animal
     are the same, ie, the order of specifying the attributes of a type
       doesn't matter.  This is not true of single inheritence systems, 
       eg, Java classes.
   abstract objects/types are available
     My apple       is an example of    Pippen apple
     Pippen apple   is a subtype of     apple
     apple          is a subtype of     fruit  
   IBML deals with terms not words
     terms are case sensitive and may incorporate spaces and punctionation
     eg, |Pippen apple| not 'pippen_apple'
         |Schrodiger's cat| not ???
   A set of basic entities types comes predefined for terms
   Entity sets are easily extended by adding new definitions
     with easy to use tools for creating and editing entity sets
   Extensive database of nameable objects and their properties
     'JFK' names a person usually called 'John F. Kennedy'
     The data entry for this preson knows
       that he was the 35th president of the United States
   terms may have many different senses.
     This and other lexical information is associated with terms
   somtware tools for language processing

a broad generalization of the more traditional Named Entity sets (e.g., The
first - defined during MUC (Grishman et al., 1996), the set developed by IREX
(Sekine et al., 2000), and the the Extended Named Entity hierarchy (Sekine
et al., 2002)).

The usual applications for Named Entity recognition include Questions and
Answering (Q&A), Information Extraction (IE), Machine Translation (MT),
Summarization and Information Retrieval (IR) and general search.


The previous Named Entiry systems have rigided type hiearchies as determined
by the designer.  The IBUKI Named Entity Hierarchy is described as a
collection of IBML types and which can be replaced by sonething other
definitions of the provided ones seem poorly chosen.  The software for
actually doing NER simply takes this collection as a parameter.

look at: http://nlp.cs.nyu.edu/ene/version6_1_0eng.html