def __init__(self):
self.__L1Tree = Parsetree()
self.__L2Tree = Parsetree()
self.__align = {}
self.__utils = Utils()
self.__phraseMap = {}
class CSHandler:
def __init__(self):
self.__L1Tree = Parsetree()
self.__L2Tree = Parsetree()
self.__align = {}
self.__utils = Utils()
self.__phraseMap = {}
def updatePhraseMap(self, phraseMap):
self.__phraseMap = phraseMap
def updateHandler(self, l1Sent, l2Sent, alignLine, l1Index):
l2Index = 1-l1Index
if l1Index:
self.__L1Tree.updateTree(l2Sent)
self.__L2Tree.updateTree(l1Sent)
else:
self.__L1Tree.updateTree(l1Sent)
self.__L2Tree.updateTree(l2Sent)
self.__align = self.__parseAlign(alignLine, l1Index, l2Index)
def __parseAlign(self, alignLine, l1Index, l2Index):
align = {}
for i in alignLine.split():
key = int(i.split("-")[l1Index])
value = int(i.split("-")[l2Index])
if key in align.keys():
align[key].append(value)
else:
align[key] = [value]
return align
def csSentence(self, csType):
validSequences = self.__utils.validSequences(self.__L1Tree.sentLen())
# Debugging !!
#sys.stderr.write("L1SeqL2Cont Valid Sequences:\n")
#for sequence in validSequences:
# sys.stderr.write(str(sequence)+"\n")
#dummy=raw_input()
###############
# Assumption that a sentence will have a single code switch.
sequence = self.__selectSequence(validSequences, csType)
# Debugging !!
#sys.stderr.write("L1SeqL2Cont Selected Sequence: "+str(sequence)+"\n")
#sys.stderr.write(l1Parse)
#sys.stderr.write(l2Parse)
#dummy=raw_input()
if sequence == -1:
return [-1,[]]
csSentence = self.__utils.makeCSSentence(self.__L1Tree.wordTags(), sequence, self.__L2Tree.wordTags(), self.__align)
return [csSentence,sequence]
## Assumptions:
## There is always a single code switch
## The selection among the valid candidate sequences is random
def __selectSequence(self, validSequences, csType):
if csType == 0:
return self.__seqL1SeqL2Cont(validSequences)
elif csType == 1:
return self.__seqL1ConstL2Cont(validSequences)
elif csType == 2:
return self.__seqL1ConstL2Const(validSequences)
elif csType == 3:
return self.__seqL1ConstL2SameConst(validSequences)
def __seqL1SeqL2Cont(self, sequences):
random.seed()
sequenceIndex = -1
l1Sequence = []
count = 0
while 1:
count += 1
if count%100 == 0:
return -1
sys.stderr.write("L1SeqL2Cont"+"InfLoop:"+str(count)+"\n")
sequenceIndex = random.randrange(len(sequences))
l1Sequence = sequences[sequenceIndex]
l2Sequence = self.__utils.l2Sequence(l1Sequence, self.__align)
if self.__L2Tree.isContiguous(l2Sequence):
# Debugging !!
#sys.stderr.write("Alignment: "+str(align)+"\n")
#sys.stderr.write("Contiguous l2Sequence: "+str(l2Sequence)+"\n")
##############
break
return l1Sequence
def __seqL1ConstL2Cont(self, sequences):
random.seed()
sequenceIndex = -1
l1Sequence = []
count = 0
while 1:
count += 1
if count%100 == 0:
return -1
sys.stderr.write("L1Const2Cont"+"InfLoop:"+str(count)+"\n")
sequenceIndex = random.randrange(len(sequences))
l1Sequence = sequences[sequenceIndex]
l2Sequence = self.__utils.l2Sequence(l1Sequence, self.__align)
if self.__L1Tree.isConstituent(l1Sequence):
if self.__L2Tree.isContiguous(l2Sequence):
# Debugging !!
#sys.stderr.write("Alignment: "+str(align)+"\n")
#sys.stderr.write("Contiguous l2Sequence: "+str(l2Sequence)+"\n")
##############
break
return l1Sequence
def __seqL1ConstL2Const(self, sequences):
random.seed()
sequenceIndex = -1
l1Sequence = []
count = 0
while 1:
count += 1
if count%100 == 0:
return -1
sys.stderr.write("L1Const2Const"+"InfLoop:"+str(count)+"\n")
sequenceIndex = random.randrange(len(sequences))
l1Sequence = sequences[sequenceIndex]
l2Sequence = self.__utils.l2Sequence(l1Sequence, self.__align)
if self.__L1Tree.isConstituent(l1Sequence):
if self.__L2Tree.isConstituent(l2Sequence):
# Debugging !!
#sys.stderr.write("Alignment: "+str(align)+"\n")
#sys.stderr.write("L2Sequence: "+str(l2Sequence)+"\n")
##############
break
return l1Sequence
def __seqL1ConstL2SameConst(self, sequences):
random.seed()
sequenceIndex = -1
l1Sequence = []
count = 0
while 1:
count += 1
if count%500 == 0:
return -1
sys.stderr.write("L1Const2SameConst"+"InfLoop:"+str(count)+"\n")
sequenceIndex = random.randrange(len(sequences))
l1Sequence = sequences[sequenceIndex]
l2Sequence = self.__utils.l2Sequence(l1Sequence, self.__align)
if self.__L1Tree.isConstituent(l1Sequence):
if self.__L2Tree.isConstituent(l2Sequence):
l1PhraseTag = self.__L1Tree.getPhrase(l1Sequence[0])
l2PhraseTag = self.__L2Tree.getPhrase(l2Sequence[0])
## Both the phrases are same, for dual structure principle
if l1PhraseTag == l2PhraseTag or l2PhraseTag in self.__phraseMap[l1PhraseTag]:
# Debugging !!
#sys.stderr.write("Alignment: "+str(align)+"\n")
#sys.stderr.write("L2Sequence: "+str(l2Sequence)+" Same Const: "+l1PhraseTag+"\n")
##############
break
return l1Sequence
