class Analyser:
wnmapping={'N':wn.NOUN,'V':wn.VERB,'J':wn.ADJ,'R':wn.ADV}
labels=['Word1_mean','Word2_mean','Cpd_mean']
def __init__(self,filepath):
self.myReader=IjcnlpReader(filepath)
def getsynlists(self,alist):
incount=0
total=0
synsetslists=[]
for token in alist:
total+=1
(lex,tag)=token.split('/')
compsynsets=wn.synsets(lex,pos=Analyser.wnmapping[tag])
if len(compsynsets)>0:
incount+=1
synsetslists.append(compsynsets)
print "Proportion of list with one entry in WN is "+str(incount)+" out of "+str(total)
return synsetslists
def checkcomposition(self,comp_synlist,word_synlist):
#check compositionality of comp with respect to word
#check over all pairings of senses - compositional if at least one sense pairing suggests compositional
result=False
for csyn in comp_synlist:
print csyn.definition
for wsyn in word_synlist:
if wsyn == csyn: #both in same synset
print "Synonyms"
return True
elif wsyn in csyn.lowest_common_hypernyms(wsyn):#check if lowest common hypernym is word
print "Word is hypernym"
return True
else:
#need to check dictionary definition too
comp_def=csyn.definition
#print comp_def
#print wsyn.lemmas,wsyn.lemma_names
for lemma in wsyn.lemmas:
tocheck= [lemma]+lemma.derivationally_related_forms()
#tocheck=[lemma]
#print tocheck
for checklemma in tocheck:
#print checklemma.name
if checklemma.name in comp_def.split(' '):
print "Found in definition: "+checklemma.name
return True
def analyse(self):
comps= self.myReader.getWNComps()
words=[self.myReader.getWNwords(1),self.myReader.getWNwords(2)]
compsynsets=self.getsynlists(comps)
wordsynsets=[self.getsynlists(words[0]),self.getsynlists(words[1])]
count=0
word1comps=[]
word2comps=[]
neithercomps=[]
bothcomps=[]
not_in_wordnet=[]
for (phrase,comp,word1,word2) in zip(comps,compsynsets,wordsynsets[0],wordsynsets[1]):
print "Checking "+phrase
if len(comp)==0:
not_in_wordnet.append(phrase)
print "Ignoring out of WN phrase: "+phrase
else:
if len(word1) ==0:
print "Warning word 1 not a noun in WN: "+phrase
if len(word2) ==0:
print "Warning word 2 not a noun in WN: "+phrase
print "Checking word 1"
w1=self.checkcomposition(comp,word1)
print "Checking word 2"
w2=self.checkcomposition(comp,word2)
#print "With respect to word1: "+str(w1)
#print "With respect to word2: "+str(w2)
if w1:
count+=1
word1comps.append(phrase)
if w2:
count+=1
word2comps.append(phrase)
bothcomps.append(phrase)
elif w2:
count+=1
word2comps.append(phrase)
else:
neithercomps.append(phrase)
print "Total score is: "+str(count)
print "BOTH:",len(bothcomps),bothcomps
self.domeans(bothcomps)
print "WORD1:",len(word1comps),word1comps
self.domeans(word1comps)
print "WORD2:",len(word2comps),word2comps
self.domeans(word2comps)
print "NEITHER:",len(neithercomps),neithercomps
self.domeans(neithercomps)
print "NOT_IN_WN:",len(not_in_wordnet),not_in_wordnet
self.domeans(not_in_wordnet)
def domeans(self,complist):
for label in Analyser.labels:
scores=self.myReader.getScores(complist,label)
#print label,len(scores),scores
array1=np.array([score for (word,score) in scores])
m=np.mean(array1)
s=np.std(array1)
e=s/math.pow(len(scores),0.5)
k1=m-e*1.96
k2=m+e*1.96 #95% confidence intervals
print label, m, s, e, '['+str(k1)+','+str(k2)+']'
def __init__(self,filepath):
self.myReader=IjcnlpReader(filepath)