def _parse_userPref(self, chktree):
if self.vb:
print "Called _parse_userPref(chktree):- EntitySet"
## what will be returned:
entSet = EntitySet("userPreference", self.vb)
entSet = self._parse_entity_pairs(chktree, entSet)
if entSet.count_entities() == 0:
## If there aren't any entity pairs
## Then perhaps this isn't a userPref utterance.
## Return Unknown for the DM.
entSet = EntitySet("unknown")
return self._expandMultiEntitySets(entSet)
def _parse_as_when(self, chktree):
if self.vb:
print "Called _parse_as_when(chktree):- EntitySet"
entSet = EntitySet("trivia", self.vb)
parameter_list = ["year"]
entSet.add_entity("type", "year")
found_subject = False
for x in chktree:
if self._isTreeBranch(x):
#if "Tree" in str(type(x)) :
if x.node == "TITLE":
found_subject = True
parameter_list.append('title')
parameter_list.append(self._getTreeTxt(x))
entSet.add_entity("movieTitle", self._getTreeTxt(x))
elif x.node == "PERSON":
found_subject = True
parameter_list.append('person')
parameter_list.append(self._getTreeTxt(x))
entSet.add_entity("person", self._getTreeTxt(x))
if found_subject == False:
parameter_list.append('other')
parameter_list.append('that')
print "_parse_as_when: append 'other' and 'that' to parameter_list"
#return parameter_list
return [ entSet ]
def _expandMultiEntitySets(self, entSet):
if self.vb:
print "before expanding entSet: %s" % (entSet.toString())
print ""
returnVar = [ entSet.copy() ]
for k,v in entSet.get_entityitems():
x = re.search(r'^(\!)?([^\d]+)(\d+)?', k)
if not returnVar[0].has_entity('rating'):
pass
if x.group(1) and returnVar[0].has_entity('rating'):
## found '!' and we have something we can inverse 'rating':
returnVar[0].remove_entity(k)
#newEntSet = returnVar[0].copy()
newEntSet = EntitySet(returnVar[0].get_classifier())
newEntSet.add_entity(x.group(2), v)
## update rating
theRating = returnVar[0].find_entity('rating')
if theRating == 5:
theRating = 1
elif theRating == 4:
theRating = 2
elif theRating == 2:
theRating = 4
elif theRating == 1:
theRating = 5
newEntSet.add_entity('rating', theRating)
returnVar.append(newEntSet)
elif x.group(3):
## numeric value found!
returnVar[0].remove_entity(k)
#newEntSet = returnVar[0].copy()
newEntSet = EntitySet(returnVar[0].get_classifier())
newEntSet.add_entity(x.group(2), v)
newEntSet.add_entity('rating', returnVar[0].find_entity('rating'))
returnVar.append(newEntSet)
return returnVar
def _parse_as_TF(self, chktree):
if self.vb:
print "Called _parse_as_TF(chktree):- EntitySet"
entSet = EntitySet("trivia_yesno", self.vb)
entSet = self._parse_entity_pairs(chktree, entSet)
return [ entSet ]
## Ignore the rest:
looking_for = "person"
negation = False
parameter_list = []
for x in chktree:
if self._isLeaf(x):
#if "tuple" in str(type(x)):
negation = self._negator(x[0], negation)
if x[1] == "CD":
parameter_list.append('year')
parameter_list.append(x[0])
entSet.add_entity("year", x[0])
elif ( x[1] == "KW_DIRECTOR" ):
looking_for = 'director'
elif ( x[1] == "KW_STAR" ):
looking_for = 'actor'
elif x[1] == "KW_PLOT":
looking_for = "plot"
elif x[1] == "GNRE":
subject = "genre"
if negation:
subject = "!genre"
parameter_list.append('!genre')
else:
parameter_list.append('genre')
parameter_list.append(x[0].title())
entSet.add_entity(subject, x[0].title())
else:
if x.node == "TITLE":
parameter_list.append('title')
parameter_list.append(self._getTreeTxt(x))
entSet.add_entity("movieTitle", self._getTreeTxt(x))
elif x.node == "PERSON":
if looking_for == "actor":
if negation:
parameter_list.append('!actor')
else:
parameter_list.append('actor')
elif looking_for == "director":
if negation:
parameter_list.append('!director')
else:
parameter_list.append('director')
else:
if negation:
parameter_list.append('!person')
else:
parameter_list.append('person')
parameter_list.append(self._getTreeTxt(x))
entSet.add_entity(looking_for, self._getTreeTxt(x))
#return parameter_list
return [ entSet ]
def _parse_as_what(self, chktree):
if self.vb:
print "Called _parse_as_what(%s):- EntitySet" % (chktree.node)
entSet = EntitySet("trivia", self.vb)
looking_for = 'other'
negation = False
## Shouldn't need this with changes to KW_PLOT below:
'''
flat = chktree.leaves()
if flat[ len(flat) - 1 ][1] == ":":
last = flat[ len(flat) - 2 ]
else:
last = flat[ len(flat) - 1 ]
if last[1] == "KW_PLOT":
entSet.add_entity("type", "plot")
'''
for itor,x in enumerate(chktree):
if self._isLeaf(x):
nodetype = self._getLeafTos(x)
nodetxt = self._getLeafTxt(x)
if not entSet.has_entity('type'):
if nodetype == "KW_YEAR":
return self._parse_as_when(chktree)
elif nodetype == "KW_DIRECTOR" or nodetype == "KW_STAR":
next = chktree[itor+1]
if self._isTreeBranch(next) and next.node == 'PERSON':
looking_for = 'actor'
if nodetype == 'KW_DIRECTOR':
looking_for = 'director'
else:
return self._parse_as_who(chktree)
elif nodetype == "KW_GENRE":
entSet.add_entity("type", "genre")
elif nodetype == "KW_MOVIE":
entSet.add_entity("type", "movieTitle")
elif nodetype == "KW_PLOT":
entSet.add_entity("type", "plot")
elif nodetype == "GNRE":
entSet.add_entity("type", "movieTitle")
entSet.add_entity("genre", nodetxt)
else:
negation = self._negator(nodetxt, negation)
if nodetype == 'POS':
## We mis-categorized the type.
## Ex: "What was director Michael Bay's latest movie?"
## Director is not what the user is looking for. director describes Michael Bay, which describes movie.
#if entSet.has_entity('type'):
# looking_for = entSet.find_entity('type')
# if looking_for = 'director' or looking_for == 'actor'
pass
if nodetype == "KW_DIRECTOR":
looking_for = "director"
elif nodetype == "KW_STAR":
looking_for = "actor"
elif nodetype == "KW_PLOT":
#looking_for = "plot"
## TODO: TEST THIS
entSet.add_entity('type', 'plot')
elif nodetype == "GNRE":
entSet.add_entity2('genre', nodetxt, negation)
negation = False
else:
if x.node == "TITLE":
entSet.add_entity2('movieTitle', self._getTreeTxt(x), negation)
elif x.node == "PERSON":
subject = looking_for
if looking_for == 'other':
subject = 'person'
entSet.add_entity2(subject, self._getTreeTxt(x), negation)
negation = False
return [ entSet ]
def _parse_as_who(self, chktree):
if self.vb:
print "Called _parse_as_who(chktree):- EntitySet"
entSet = EntitySet("trivia", self.vb)
negation = False
looking_for = 'person'
for x in chktree:
if self._isLeaf(x):
negation = self._negator(self._getLeafTxt(x), negation)
keyword = None
if self._getLeafTos(x) == "KW_DIRECTOR":
keyword = 'director'
elif self._getLeafTos(x) == "KW_STAR":
keyword = 'actor'
if keyword:
if not entSet.has_entity('type'):
entSet.add_entity('type', keyword)
else:
looking_for = keyword
else:
if x.node == "TITLE":
entSet.add_entity2("movieTitle", self._getTreeTxt(x), negation)
negation = False
elif x.node == "PERSON":
personName = self._getTreeTxt(x)
if not entSet.has_entity('type'):
entSet.add_entity('type', 'person')
entSet.add_entity('character', personName)
else:
entSet.add_entity2(looking_for, personName, negation)
negation = False
if entSet.count_entities() == 0:
#empty entitySet -> unable to parse utterance -> let DM know
entSet.change_classifier("unknown")
elif not entSet.has_entity('type'):
entSet.add_entity('type', 'person')
return [ entSet ]
def _parse_command(self, chunked):
if self.vb:
print "Called _parse_command(chktree):- EntitySet"
entSet = False
'''
cmd_node = chunked[0]
verbs = self._find_verbs(cmd_node)
theVerb = None
if len(verbs) == 1:
theVerb = verbs[0].lower()
if theVerb == 'know':
#I would like to know directed "Twilight"?
#know the cast of "Ocean's 11"?
#Do you know which movies Keanu Reeves is in?
return self._rechunk(chunked)
else:
theVerb = verbs[0]
#theClassifier = self.verbToClassifierMapping.get(verbs[0].lower(), "unknown")
elif len(verbs) == 0:
# Just in Case Error-Catch
if vb.self:
print "NLU Error: Parsing COMMAND node resulted in no VB node. Impossible; grammar does not allow this"
pass
else:
#TODO: Will this happen? (will there ever be more than one verb in command node?
# if so, program!
#theClassifier = self.verbToClassifierMapping.get(verbs[0].lower(), "unknown")
theVerb = verbs[0]
'''
theClassifier = None
theVerb = None #debugging var
score = 0.0
negator = False
for x in chunked[0]:
#ensure this isn't an adverb
if self._isLeaf(x) and self._getLeafTos(x)[0] == 'V':
theClassifier,score = self._classify_word(x, self.classifierSS_list)
theVerb = self._getLeafTxt(x)
negator = self._negator(theVerb, negator)
break
#elif self._isTreeBranch(x):
#for y in x:
#if self._isLeaf(y):
#negator = self._negator(y)
#theClassifier,score = self._classify_word(theVerb, self.classifierSS_list)
if self.vb:
print "parse_cmd: %s -> %s with score %.05f" % (theVerb, theClassifier, score)
if theClassifier == None or score < .25:
theClassifier = "unknown"
type = False
for itor,x in enumerate(chunked):
if itor < 1:
continue
if self._isTreeBranch(x): ## this shouldn't happen, but just in case
continue
tagOfSpeech = self._getLeafTos(x)
theWord = self._getLeafTxt(x)
negator = self._negator(theWord, negator)
## stage look-forward or look-backward:
keyword = ""
if tagOfSpeech == 'KW_DIRECTOR':
type = "director"
elif tagOfSpeech == 'KW_STAR':
type = 'actor'
elif tagOfSpeech == 'KW_GENRE':
type = 'genre'
elif tagOfSpeech == 'KW_PLOT':
type = 'plot'
theClassifier = 'trivia'
elif tagOfSpeech == 'KW_MOVIE' or tagOfSpeech == 'GNRE':
## include GNRE to catch sentences, such as "I want to watch a Comedy"
type = 'movieTitle'
itorBackTrack = itor
while itorBackTrack > 0:
y = chunked[itorBackTrack]
if self._isTreeBranch(y):
break
yTos = self._getLeafTos(y)
if yTos.startswith('JJ') or yTos.startswith('NN'):
adj, score = self._classify_word(y, wn.synsets('recent'))
if score > .25:
entSet.add_entity('timespan', 'recent')
else:
adj, score = self._classify_(y, wn.synset('old'))
#'recent'
#'old'
#'new'
#'first'
#'last'
#'popular'
#'similar'
#'more'
else:
break
if type:
break
entSet = EntitySet(theClassifier)
entSet.add_entity("type", type)
entSet = self._parse_entity_pairs(chunked, entSet)
entSet.remove_entity('rating')
return [ entSet ]