def generateName(self):
name = INITIAL_CHAR
probabilityOfName = 1.0
while len(name) < self.maxNameLength_ + 1:
minNgramLength, maxNGramLength = self.prefixCounter_.rangeTuple()
#Do not try to have prefixes longer that what was already generated
maxNGramLength = min(maxNGramLength, len(name))
#Compute weights to choose next character
for length in range(minNgramLength, maxNGramLength + 1):
prefix = name[-length:] if length > 0 else ''
totalOccurences = self.prefixCounter_.countPrefixOccurences(
length, prefix)
perCharOccurences = list(
self.prefixCounter_.perCharacterPrefixOccurences(
length, prefix))
for char, charOccurences in perCharOccurences:
self.scorer_.learn(length, char, totalOccurences,
charOccurences)
transitionCharacters, transitionScores = map(
list, zip(*self.scorer_.scores()))
#Discard the possibility to finish the name if not enough characters were generated.
if len(name) < self.minNameLength_:
if TERMINAL_CHAR in transitionCharacters:
i = transitionCharacters.index(TERMINAL_CHAR)
del transitionCharacters[i]
del transitionScores[i]
#If there is no meaningful character to follow the prefix, raise an exception
#( Should never happen if `minNGramLength == 0`, since then the empty prefix is used and any character
#that appears in the lexicon can be picked with at least very unlikely prabilities.
if len(transitionCharacters) <= 0:
namePerplexity = perplx(probabilityOfName, len(name))
raise InvalidGeneratedWord(
("No character to transition from \"%s\"" % prefix), name,
namePerplexity)
#Pick nexct character and build name
i = discretepick(transitionScores)
character = transitionCharacters[i]
if character == TERMINAL_CHAR:
break
name += character
probabilityOfName *= transitionScores[i] / sum(transitionScores)
self.scorer_.reset()
name = name[1:]
namePerplexity = perplx(probabilityOfName, len(name))
return (namePerplexity, name)
def generateName( self ):
name = INITIAL_CHAR
probabilityOfName = 1.0
while len(name) < self.maxNameLength_ + 1:
minNgramLength, maxNGramLength = self.prefixCounter_.rangeTuple()
#Do not try to have prefixes longer that what was already generated
maxNGramLength = min( maxNGramLength, len(name) )
#Compute weights to choose next character
for length in range(minNgramLength,maxNGramLength+1):
prefix = name[-length:] if length > 0 else ''
totalOccurences = self.prefixCounter_.countPrefixOccurences( length, prefix )
perCharOccurences = list( self.prefixCounter_.perCharacterPrefixOccurences( length, prefix ) )
for char, charOccurences in perCharOccurences:
self.scorer_.learn( length, char, totalOccurences, charOccurences )
transitionCharacters, transitionScores = map( list,zip( *self.scorer_.scores() ) )
#Discard the possibility to finish the name if not enough characters were generated.
if len(name) < self.minNameLength_:
if TERMINAL_CHAR in transitionCharacters:
i = transitionCharacters.index( TERMINAL_CHAR )
del transitionCharacters[i]
del transitionScores[i]
#If there is no meaningful character to follow the prefix, raise an exception
#( Should never happen if `minNGramLength == 0`, since then the empty prefix is used and any character
#that appears in the lexicon can be picked with at least very unlikely prabilities.
if len(transitionCharacters) <= 0:
namePerplexity = perplx( probabilityOfName, len(name) )
raise InvalidGeneratedWord( ("No character to transition from \"%s\""%prefix), name, namePerplexity )
#Pick nexct character and build name
i = discretepick( transitionScores )
character = transitionCharacters[i]
if character == TERMINAL_CHAR:
break
name += character
probabilityOfName *= transitionScores[i] / sum(transitionScores)
self.scorer_.reset()
name=name[1:]
namePerplexity = perplx( probabilityOfName, len(name) )
return ( namePerplexity, name )
def generateName(self):
name = INITIAL_CHAR
probabilityOfName = 1.0
while len(name) < self.maxNameLength_ + 1:
# Compute weights to choose next character
prefix = name[-self.nGramLength_ :] if self.nGramLength_ > 0 else ""
transitions = list(self.transitionTable_.transitionsForPrefix(prefix))
# If there is no meaningful character to follow the prefix, raise an exception
if len(transitions) <= 0:
namePerplexity = perplx(probabilityOfName, len(name))
raise InvalidGeneratedWord(('No character to transition from "%s"' % prefix), name, namePerplexity)
transitionCharacters, transitionScores = map(list, zip(*transitions))
# Discard the possibility to finish the name if not enough characters were generated.
if len(name) < self.minNameLength_:
if TERMINAL_CHAR in transitionCharacters:
i = transitionCharacters.index(TERMINAL_CHAR)
del transitionCharacters[i]
del transitionScores[i]
if len(transitionCharacters) <= 0:
namePerplexity = perplx(probabilityOfName, len(name))
raise InvalidGeneratedWord(
('Only end-of-word transition from "%s"' % prefix), name, namePerplexity
)
i = discretepick(transitionScores)
character = transitionCharacters[i]
# Pick nexct character and build name
if character == TERMINAL_CHAR:
break
name += character
probabilityOfName *= transitionScores[i] / sum(transitionScores)
name = name[1:]
namePerplexity = perplx(probabilityOfName, len(name))
return (namePerplexity, name)
def generateName( self ):
name = INITIAL_CHAR
probabilityOfName = 1.0
while len(name) < self.maxNameLength_ + 1:
#Compute weights to choose next character
prefix = name[-self.nGramLength_:] if self.nGramLength_ > 0 else ''
transitions = list( self.transitionTable_.transitionsForPrefix( prefix ) )
#If there is no meaningful character to follow the prefix, raise an exception
if len(transitions) <= 0:
namePerplexity = perplx( probabilityOfName, len(name) )
raise InvalidGeneratedWord( ("No character to transition from \"%s\""%prefix), name, namePerplexity )
transitionCharacters, transitionScores = map( list, zip( *transitions ) )
#Discard the possibility to finish the name if not enough characters were generated.
if len(name) < self.minNameLength_:
if TERMINAL_CHAR in transitionCharacters:
i = transitionCharacters.index( TERMINAL_CHAR )
del transitionCharacters[i]
del transitionScores[i]
if len(transitionCharacters) <= 0:
namePerplexity = perplx( probabilityOfName, len(name) )
raise InvalidGeneratedWord( ("Only end-of-word transition from \"%s\""%prefix), name, namePerplexity )
i = discretepick( transitionScores )
character = transitionCharacters[i]
#Pick nexct character and build name
if character == TERMINAL_CHAR:
break
name += character
probabilityOfName *= transitionScores[i] / sum(transitionScores)
name=name[1:]
namePerplexity = perplx( probabilityOfName, len(name) )
return (namePerplexity, name)