#  ---- Recognizer Pattern File ----

# from pattern.h:

#  Possible Pattern Predicates:
#     <NP>
#     <NP-SUBJ>
#     <NP-DOBJ>
#     <NP-IOBJ>
#     <NP-DEF>
#     <NP-HEAD:sem=semanticfeaturehere>
#     <NP-HEAD:syn=syntacticfeaturehere>
#     <NP-HEAD:tag=syntactictaghere>
#     <NP-HEAD:word=wordvaluehere>
#     <NP-ANYWORD:sem=semanticfeaturehere>
#     <NP-ANYWORD:syn=syntacticfeaturehere>
#     <NP-ANYWORD:tag=syntactictaghere>
#     <NP-ANYWORD:word=wordvaluehere>
#     <VP>
#     <VP-ACTIVE>
#     <VP-PASSIVE>
#     <VP-INF>
#     <VP-SIMPLE-INF>   (only "to <verb>")
#     <PP>
#     <WORD:word=wordvaluehere>
#     <WORD:sem=semanticfeaturehere>
#     <WORD:tag=syntactictaghere>
#     <PUNC:word=punctuationtypehere>
#
#  Combinations of predicates are allowed and considered to be
#  AND'ed together.
#     
#  All predicates should be followed by "->" and then the type
#  of constituent constituent you want to be created followed
#  by a dash and the pattern type: 
#    NP-LOCATION,
#    NP-COMPANY,
#    NP-NUMBER,
#    NP-MONEY,
#    NP-PERCENT,
#    NP-TIME
#    NP-ING.
#
#
#  NOTE: Because of the way that ApplyConstPattern() works, every
#  predicate must target a constituent of the type that would exist
#  after sundance segmenting at the point that all this pattern
#  recognition gets called.  For example, if we call this after np
#  segmenting, then the VP and PP predicates really don't do anything
#  for us.  Also, since each predicate must target one constituent
#  in a sequence, there is no facility for looking at, for example,
#  two particular words within the same NP.  If we ever want to change
#  this, we've got to make the necessary changes in the ApplyConstPattern()
#  function.  Note also that since the ApplyConstPattern() function
#  returns pairs of numbers which equate to the starting and ending
#  position of a sentence's children which should be rolled together,
#  this recognizer code cannot be called after clause handling because
#  clause handling creates a 3-level parse and we've just assumed
#  a 2-level parse.
#
#  So, because the current implementation applies these patterns
#  after NP segmenting, but before any other segmenting, the examples
#  with VP's and PP's are irrelevant for now.  (The logic is there in
#  sundance, we'd just need to have multiple pattern files to be
#  called in at different points of the parsing.)
#
#  Pattern Precedence Rules : The pattern recognizer now enforces a
#  type of precedence when more than one pattern may apply to
#  a set of shared constituents.  The rule is simple...the longest
#  one wins, and in the case of ties, the rule that is listed first
#  in the rule file wins.  

#
#Part added by Pol Schumacher 
#
#This pattern should recognize labels which are in "
#<WORD:tag=Verb> <WORD:word=tab> -> NP-TIME
#<WORD:case=Capitalized> <WORD:word=button> -> NP-TIME
#Pattern for keys

<WORD:word=crtl><WORD:word=a> -> NP-LIST
<WORD:word=alt><WORD:word=a> -> NP-LIST
<WORD:word=shift><WORD:word=a> -> NP-LIST

#Menu tree structures
<PUNC:word=GREATER><NP> ->NP-LIST
<PUNC:word=GREATER><NP><PUNC:word=GREATER><NP> -> NP-LIST
<PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP> -> NP-LIST
<PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP> -> NP-LIST
<PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP> -> NP-LIST
<PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP> -> NP-LIST
<PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP><PUNC:word=GREATER><NP> -> NP-LIST
#Pattern for time
#<WORD:word=for> <WORD:tag=NUMBER> <WORD:word=to> <WORD:tag=NUMBER> <WORD:word=minutes> -> NP-TIME
#<WORD:word=for> <WORD:tag=NUMBER><PUNC:word=MINUS><WORD:tag=NUMBER> <WORD:word=minutes> -> NP-TIME
#<WORD:tag=NUMBER><PUNC:word=MINUS><WORD:tag=NUMBER> <WORD:word=min.> -> NP-TIME
#<WORD:tag=NUMBER><WORD:word=to><WORD:tag=NUMBER> <WORD:word=minutes> -> NP-TIME
#<WORD:tag=NUMBER> <WORD:word=min> -> NP-TIME
#<WORD:tag=NUMBER> <WORD:word=minutes> -> NP-TIME
#<WORD:tag=NUMBER> <WORD:word=min.> -> NP-TIME
#<WORD:tag=NUMBER><PUNC:word=MINUS><WORD:tag=NUMBER> <WORD:word=min.> -> NP-TIME
#<WORD:tag=NUMBER><WORD:word=to><WORD:tag=NUMBER> <WORD:word=minutes> -> NP-TIME
##Pattern buttons
<PUNC:word=SQUOTE><NP><PUNC:word=SQUOTE> -> NP-LIST

##Pattern for ingredient list
<NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA><NP> <PUNC:word=COMMA> <NP> -> NP-LIST


##Pattern comma and
<NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST

##Pattern comma or
<NP> <PUNC:word=COMMA> <NP> <WORD:word=and> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=or> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=or> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=or> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=or> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=or> <NP> -> NP-LIST
<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP> <WORD:word=or> <NP> -> NP-LIST


#End Pol's part

# To capture "IBM Corp." "Nestle Inc." and "L. L. Bean Co."
###<NP-HEAD:word=Corp.> -> NP-COMPANY
###<NP-HEAD:word=Inc.> -> NP-COMPANY
###<NP-HEAD:word=Co.> -> NP-COMPANY

# To capture : "Canada"
###<NP-HEAD:sem=COUNTRY> -> NP-LOCATION

# To capture : Alberta, Canada
<NP> <PUNC:word=COMMA> <NP-HEAD:sem=COUNTRY> -> NP-LOCATION
# NOTE: can't do this!!! Really need for *contextual* features in
# these rules! The preposition "in" is a crucial part of recognizing
# that Ames is a location and therefore a  necessary part of the
# pattern, but it shouldn't be pulled into the NP itself!
#
# Ex: "in Ames, Iowa"
# where Ames is an unknown word but Iowa is a known LOCATION.
# I think this should be a good and pretty safe rule -emr 8/3/07
#
# <WORD:word=IN> <NP> <PUNC:word=COMMA> <NP-HEAD:sem=LOCATION> -> NP-LOCATION

# To capture : Calgary, Alberta, Canada
###<NP> <PUNC:word=COMMA> <NP> <PUNC:word=COMMA> <NP-HEAD:sem=COUNTRY> -> NP-LOCATION

#===================
# More aggressive location tagging...
<NP-HEAD:sem=LOCATION> -> NP-LOCATION

# Date/time tagging--anyword or head
###<NP-ANYWORD:sem=TIME> -> NP-TIME
<NP-HEAD:sem=TIME> -> NP-TIME

#<WORD:word=TODAY> -> NP-TIME
#<WORD:word=YESTERDAY> -> NP-TIME
#<WORD:word=TOMORROW> -> NP-TIME

#<WORD:word=MONDAY> -> NP-TIME
#<WORD:word=TUESDAY> -> NP-TIME
#<WORD:word=WEDNESDAY> -> NP-TIME
#<WORD:word=THURSDAY> -> NP-TIME
#<WORD:word=FRIDAY> -> NP-TIME
#<WORD:word=SATURDAY> -> NP-TIME
#<WORD:word=SUNDAY> -> NP-TIME
<WORD:tag=NUMBER> <WORD:word=p.m.> -> NP-TIME
<WORD:tag=NUMBER> <WORD:word=a.m.> -> NP-TIME

<WORD:sem=MONTH> <WORD:tag=NUMBER> <PUNC:word=COMMA> <WORD:tag=NUMBER> -> NP-TIME

# added rule below because some years are labeled as TIMEs in the dictionary -emr
<WORD:sem=MONTH> <WORD:tag=NUMBER> <PUNC:word=COMMA> <WORD:sem=TIME> -> NP-TIME

<WORD:sem=MONTH> <WORD:tag=NUMBER> -> NP-TIME

#<WORD:word=DOLLAR> <WORD:tag=NUMBER> <WORD:sem=DENOMINATION> -> NP-MONEY
#<WORD:word=DOLLAR> <WORD:tag=NUMBER> -> NP-MONEY

#<NP-ALLWORDS:case=Capitalized> -> NP-PROPERNOUN

# this should be subsumed by more general pattern below -emr
# <WORD:word=last> <WORD:word=night> -> NP-TIME

# This will force "May" to be parsed as a NOUN (Month) when it is
# capitalized.  A bit risky in that it will be labeled a TIME when it
# is a person name, or in a headline, or all caps.
# BUT ... realized this will mess up everything in MUC-4 corpus 'cause
# it is all caps. Grrr... -emr
# <WORD:word=May&WORD:case=Capitalized> -> NP-TIME


# ***********************************************************
#         Added cases below for VIN domain -emr
# ***********************************************************

# Ex: ## days ago ; a few days ago ; one day ago
<WORD:tag=NUMBER> <WORD:sem=TIME> <WORD:word=AGO> -> NP-TIME
<WORD:sem=TIME> <WORD:word=AGO> -> NP-TIME

# Ex: a month later
<WORD:tag=ARTICLE> <WORD:sem=TIME> <WORD:word=LATER> -> NP-TIME
<WORD:tag=NUMBER> <WORD:sem=TIME> <WORD:word=LATER> -> NP-TIME

# Ex: last week, next week, past week
<WORD:word=LAST> <WORD:sem=TIME> -> NP-TIME
<WORD:word=NEXT> <WORD:sem=TIME> -> NP-TIME
<WORD:word=PAST> <WORD:sem=TIME> -> NP-TIME

# Ex: that evening     EXPERIMENTAL! This might be too dangerous,
# though it *seems* relatively safe. -emr
<WORD:word=THAT> <WORD:sem=TIME> -> NP-TIME