def computeStatsOld(self,
absList,
statOut=None,
errorOut=None,
typeList=[]):
""" compute RPF stats for associated mentions and quantities in a list
of abstracts.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
statOut.copy(self.statList)
# how many of the associations are correct/incorrect?
componentIncorrect = 0
statDescription = '(G,O) - OM'
stats = EntityStats([statDescription])
for abstract in self.associationList.keys():
errorOut.write('---%s ---\n' % (abstract.id))
for omAssociation in self.associationList[abstract]:
group = omAssociation.group
outcome = omAssociation.outcome
er = omAssociation.outcomeMeasurement.getTextEventRate()
on = omAssociation.outcomeMeasurement.getOutcomeNumber()
errorMsgs = []
self.checkQuantity(er, group, outcome, errorMsgs)
self.checkQuantity(on, group, outcome, errorMsgs)
if len(errorMsgs) == 0:
stats.irstats[statDescription].incTP()
prefix = '+TP'
else:
stats.irstats[statDescription].incFP()
prefix = '-FP'
if (er == None or er.isTruePositive()) and (on == None or on.isTruePositive()) \
and (er == None or on == None or on.shouldBelongToSameOutcomeMeasurement(er)):
# this outcome measurement is correct and should be associated
# this was a false negative and a false positive
stats.irstats[statDescription].incFN()
errorOut.write('-FN: %s\n' %
(omAssociation.getValueString()))
errorOut.write('%s: %s\n' % (prefix, omAssociation.toString()))
if len(errorMsgs) > 0:
errorOut.write(', '.join(errorMsgs) + '\n')
# look at those OMs that were not associated
if abstract in self.incompleteMatches:
self.__processIncompleteMatches(abstract, errorOut, stats,
statDescription)
for abstract in absList:
if abstract not in self.associationList.keys():
errorOut.write('---%s ---\n' % (abstract.id))
if abstract in self.incompleteMatches:
self.__processIncompleteMatches(abstract, errorOut, stats,
statDescription)
stats.saveStats(statOut, keyPrefix='Assoc ')
def computeStats(self, absList, statOut=None, errorOut=None, typeList=[], keyPrefix=''):
""" compute RPF stats for associated mentions and quantities in a list
of abstracts.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
if len(keyPrefix) == 0:
keyPrefix = 'Assoc - '
if len(typeList) > 0:
statDescription = '-'.join(typeList)
else:
statDescription = self.entityTypesString
stats = EntityStats([statDescription])
totalFalsePairs = 0
for abs in absList:
errorOut.write('---%s (%s)\n'%(abs.id, statDescription))
for s in abs.sentences:
[tp, fp, fn, falsePairs] = self.checkAssociations(s, errorOut, typeList)
errorOut.write('tp: %d, fp: %d, fn: %d, falsePairs: %d\n'%(tp, fp, fn, falsePairs))
totalFalsePairs += falsePairs
stats.irstats[statDescription].addTP(tp)
stats.irstats[statDescription].addFP(fp)
stats.irstats[statDescription].addFN(fn)
stats.printStats()
print 'Total false pairs: ', totalFalsePairs
# stats.writeStats(statOut)
stats.saveStats(statOut, keyPrefix=keyPrefix)
return stats
def computeStats(self, absList, statOut=None, errorOut=None):
""" compute RPF stats for detected quantities in a list of abstracts.
write results to output stream.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
stats = EntityStats(self.entityTypes)
for abs in absList:
errorOut.write('---'+abs.id+'---\n')
for sentence in abs.sentences:
# decide if should output sentence
printedSentence = False
# count correct and incorrect labelings
for eType in self.entityTypes:
for token in sentence:
if self.isImportantNumber(token) == True:
if token.hasLabel(eType) and token.hasAnnotation(eType):
stats.irstats[eType].incTP()
errorOut.write('+TP: ' + token.text + ' ('+eType+')\n')
elif token.hasLabel(eType) == True:
stats.irstats[eType].incFP()
errorOut.write('-FP: ' + token.text + ' ('+eType+')\n')
elif token.hasAnnotation(eType) == True:
stats.irstats[eType].incFN()
errorOut.write('-FN: ' + token.text + ' ('+eType+')\n')
stats.printStats()
# stats.writeStats(statOut)
if statOut != None:
stats.saveStats(statOut, keyPrefix='NF - ')
return stats
def computeStatsOld(self, absList, statOut=None, errorOut=None, typeList=[]):
""" compute RPF stats for associated mentions and quantities in a list
of abstracts.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
statOut.copy(self.statList)
# how many of the associations are correct/incorrect?
componentIncorrect = 0
statDescription = '(G,O) - OM'
stats = EntityStats([statDescription])
for abstract in self.associationList.keys():
errorOut.write('---%s ---\n'%(abstract.id))
for omAssociation in self.associationList[abstract]:
group = omAssociation.group
outcome = omAssociation.outcome
er = omAssociation.outcomeMeasurement.getTextEventRate()
on = omAssociation.outcomeMeasurement.getOutcomeNumber()
errorMsgs = []
self.checkQuantity(er, group, outcome, errorMsgs)
self.checkQuantity(on, group, outcome, errorMsgs)
if len(errorMsgs) == 0:
stats.irstats[statDescription].incTP()
prefix = '+TP'
else:
stats.irstats[statDescription].incFP()
prefix = '-FP'
if (er == None or er.isTruePositive()) and (on == None or on.isTruePositive()) \
and (er == None or on == None or on.shouldBelongToSameOutcomeMeasurement(er)):
# this outcome measurement is correct and should be associated
# this was a false negative and a false positive
stats.irstats[statDescription].incFN()
errorOut.write('-FN: %s\n'%(omAssociation.getValueString()))
errorOut.write('%s: %s\n'%(prefix, omAssociation.toString()))
if len(errorMsgs) > 0:
errorOut.write(', '.join(errorMsgs)+'\n')
# look at those OMs that were not associated
if abstract in self.incompleteMatches:
self.__processIncompleteMatches(abstract, errorOut, stats, statDescription)
for abstract in absList:
if abstract not in self.associationList.keys():
errorOut.write('---%s ---\n'%(abstract.id))
if abstract in self.incompleteMatches:
self.__processIncompleteMatches(abstract, errorOut, stats, statDescription)
stats.saveStats(statOut, keyPrefix='Assoc ')
def computeStats(self,
absList,
statOut=None,
errorOut=None,
typeList=[],
keyPrefix=''):
""" compute RPF stats for associated mentions and quantities in a list
of abstracts.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
if len(keyPrefix) == 0:
keyPrefix = 'Assoc - '
if len(typeList) > 0:
statDescription = '-'.join(typeList)
else:
statDescription = self.entityTypesString
stats = EntityStats([statDescription])
totalFalsePairs = 0
for abs in absList:
errorOut.write('---%s (%s)\n' % (abs.id, statDescription))
for s in abs.sentences:
[tp, fp, fn,
falsePairs] = self.checkAssociations(s, errorOut, typeList)
errorOut.write('tp: %d, fp: %d, fn: %d, falsePairs: %d\n' %
(tp, fp, fn, falsePairs))
totalFalsePairs += falsePairs
stats.irstats[statDescription].addTP(tp)
stats.irstats[statDescription].addFP(fp)
stats.irstats[statDescription].addFN(fn)
stats.printStats()
print 'Total false pairs: ', totalFalsePairs
# stats.writeStats(statOut)
stats.saveStats(statOut, keyPrefix=keyPrefix)
return stats
def computeStats(self, absList, statOut=None, errorOut=None):
""" compute RPF stats for detected mentions in a list of abstracts.
write results to output stream.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
stats = EntityStats(self.entityTypes)
for abs in absList:
errorOut.write('---'+abs.id+'---\n')
# identify ALL annotated mentions, even in sentences we are not focused on
# for sentence in abs.allSentences():
# for mType in self.entityTypes:
# aList = sentence.getAnnotatedMentions(mType, recomputeMentions=True)
#
# for sentence in abs.sentences:
# for mType in self.entityTypes:
# self.compareAnnotatedAndDetected(sentence, mType, \
# stats.irstats[mType], errorOut)
for sentence in abs.allSentences():
for mType in self.entityTypes:
if sentence in abs.sentences:
self.compareAnnotatedAndDetected(sentence, mType, \
stats.irstats[mType], errorOut)
else:
aList = sentence.getAnnotatedMentions(mType, recomputeMentions=True)
stats.printStats()
if statOut != None:
stats.saveStats(statOut, keyPrefix='MF - ')
return stats
def computeStats(self, absList, statOut=None, errorOut=None, typeList=[]):
""" compute RPF stats for associated outcome measurements and group and outcome mentions in a list
of abstracts.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
statOut.copy(self.statList)
# how many of the associations are correct/incorrect?
statDescription = '(G,O) - OM'
stats = EntityStats([statDescription])
taFile = open('trueassociations.%s.txt' % self.finderType, 'w')
for abstract in absList:
errorOut.write('---%s ---\n' % (abstract.id))
trueAssociations = self.buildTrueAssociations(abstract)
taFile.write('---%s ---\n' % (abstract.id))
for ta in trueAssociations:
taFile.write(ta.toString() + '\n')
if abstract in self.associationList:
for omAssociation in self.associationList[abstract]:
group = omAssociation.outcomeMeasurement.getGroup()
outcome = omAssociation.outcomeMeasurement.getOutcome()
er = omAssociation.outcomeMeasurement.getTextEventRate()
on = omAssociation.outcomeMeasurement.getOutcomeNumber()
errorMsgs = []
matchFound = False
for ta in trueAssociations:
if ta.match == None and (
(er != None and er == ta.eventrate) or
(on != None and on == ta.outcomeNumber)):
# we have a potential match. check to see if everything matches
errorMsgs = []
self.checkQuantity(er, group, outcome, errorMsgs)
self.checkQuantity(on, group, outcome, errorMsgs)
if len(errorMsgs) == 0:
matchFound = True
ta.match = omAssociation
break
if matchFound:
stats.irstats[statDescription].incTP()
prefix = '+TP'
else:
# we have a false positive
stats.irstats[statDescription].incFP()
prefix = '-FP'
errorOut.write('%s: %s\n' %
(prefix, omAssociation.toString()))
if len(errorMsgs) > 0:
errorOut.write(', '.join(errorMsgs) + '\n')
# look at those OMs that were not associated
if abstract in self.incompleteMatches:
self.__processIncompleteMatches(abstract, errorOut)
# count false negatives
for ta in trueAssociations:
if ta.match == None:
stats.irstats[statDescription].incFN()
errorOut.write('-FN: %s\n' % (ta.toString()))
taFile.close()
stats.saveStats(statOut, keyPrefix='Assoc ')
def computeStats(self, absList, statOut=None, errorOut=None, typeList=[]):
""" compute RPF stats for associated outcome measurements and group and outcome mentions in a list
of abstracts.
write final RPF stats to statOut
write TP/FP/FN to errorOut
"""
statOut.copy(self.statList)
# how many of the associations are correct/incorrect?
statDescription = '(G,O) - OM'
stats = EntityStats([statDescription])
taFile = open('trueassociations.%s.txt'%self.finderType, 'w')
for abstract in absList:
errorOut.write('---%s ---\n'%(abstract.id))
trueAssociations = self.buildTrueAssociations(abstract)
taFile.write('---%s ---\n'%(abstract.id))
for ta in trueAssociations:
taFile.write(ta.toString() + '\n')
if abstract in self.associationList:
for omAssociation in self.associationList[abstract]:
group = omAssociation.outcomeMeasurement.getGroup()
outcome = omAssociation.outcomeMeasurement.getOutcome()
er = omAssociation.outcomeMeasurement.getTextEventRate()
on = omAssociation.outcomeMeasurement.getOutcomeNumber()
errorMsgs = []
matchFound = False
for ta in trueAssociations:
if ta.match == None and ((er != None and er == ta.eventrate) or (on != None and on == ta.outcomeNumber)):
# we have a potential match. check to see if everything matches
errorMsgs = []
self.checkQuantity(er, group, outcome, errorMsgs)
self.checkQuantity(on, group, outcome, errorMsgs)
if len(errorMsgs) == 0:
matchFound = True
ta.match = omAssociation
break
if matchFound:
stats.irstats[statDescription].incTP()
prefix = '+TP'
else:
# we have a false positive
stats.irstats[statDescription].incFP()
prefix = '-FP'
errorOut.write('%s: %s\n'%(prefix, omAssociation.toString()))
if len(errorMsgs) > 0:
errorOut.write(', '.join(errorMsgs)+'\n')
# look at those OMs that were not associated
if abstract in self.incompleteMatches:
self.__processIncompleteMatches(abstract, errorOut)
# count false negatives
for ta in trueAssociations:
if ta.match == None:
stats.irstats[statDescription].incFN()
errorOut.write('-FN: %s\n'%(ta.toString()))
taFile.close()
stats.saveStats(statOut, keyPrefix='Assoc ')
