def __init__(self, abstract, useAnnotated=False, useTrialReports=True):
""" create a list of groups and common medical conditions given
an abstact object. """
self.abstract = abstract
self.useTrialReports = useTrialReports
self.nTrueGroups = 0
self.nTrueConditions = 0
self.nTrueGroupSizes = 0
if useAnnotated == True:
self.groupTemplates = abstract.annotatedEntities.getList('group')
self.ageTemplates = createAnnotatedMergedList(abstract, 'age')
self.conditionTemplates = abstract.annotatedEntities.getList('condition')
populationTemplates = createAnnotatedMergedList(abstract, 'population')
else:
self.groupTemplates = abstract.entities.getList('group')
self.ageTemplates = createMergedList(abstract, 'age')
self.conditionTemplates = abstract.entities.getList('condition')
populationTemplates = createMergedList(abstract, 'population')
self.ageInfo = AgeInfo(self.ageTemplates, abstract, self.useTrialReports)
self.gender = Gender(populationTemplates, abstract, self.useTrialReports)
# filter useless population terms
self.populationTemplates = []
for pTemplate in populationTemplates:
if pTemplate.isInteresting() > 0:
# term is informative, keep it
self.populationTemplates.append(pTemplate)
def computeStatistics(self, errorOut):
""" Count RPF statistics for each unique AGE, CONDITION, POPULATION entity
statOut = file stream for RPF stats for all parts of summarization system
errorOut = file stream for TPs, FPs, FNs
return hash of IRstats, one for each mention type, keyed by mention type
"""
stats = {}
self.nTrueGroupSizes = 0
aAgeTemplates = createAnnotatedMergedList(self.abstract, 'age')
errorOut.write('age:\n')
stats['age'] = self.ageInfo.countAgeMatches(aAgeTemplates, errorOut)
errorOut.write('condition:\n')
aConditionTemplates = self.abstract.annotatedEntities.getList('condition')
stats['condition'] = countMatches(aConditionTemplates, \
self.conditionTemplates, errorOut)
errorOut.write('group:\n')
aGroupTemplates = self.abstract.annotatedEntities.getList('group')
stats['group'] = countMatches(aGroupTemplates, self.groupTemplates, errorOut)
self.nTrueConditions = len(aConditionTemplates)
self.nTrueGroups = len(aGroupTemplates)
errorOut.write('group size:\n')
gsStats = IRstats()
gsFound = set([])
for gTemplate in self.groupTemplates:
gSize = gTemplate.getSize(maxSize=True)
if gSize != 0:
# look for group size match in sizes for annotated group
found = False
if gTemplate.matchedTemplate != None:
for trueGSize in gTemplate.matchedTemplate.sizes:
if gSize == trueGSize.value:
found = True
break
if found:
# group size is correct
gsStats.incTP()
errorOut.write(' +TP: %s size = %d\n' % (gTemplate.name, gSize))
gTemplate.groupSizeEvaluation.markCorrect()
gsFound.add(gTemplate.matchedTemplate)
else:
# group size is incorrect
gsStats.incFP()
errorOut.write(' -FP: %s size = %d\n' % (gTemplate.name, gSize))
gTemplate.groupSizeEvaluation.markIncorrect()
# look for false negatives
for trueTemplate in aGroupTemplates:
if trueTemplate not in gsFound and trueTemplate.matchedTemplate != None and trueTemplate.getSize() > 0:
# there should be a group size for this group
gsStats.incFN()
errorOut.write(' -FN: %s size = %d\n' % \
(trueTemplate.name, trueTemplate.getSize()))
stats['group size'] = gsStats
self.nTrueGroupSizes = gsStats.tp + gsStats.fn
# errorOut.write('population:\n')
# templates = createAnnotatedMergedList(self.abstract, 'population')
# aPopulationTemplates = []
# for pTemplate in templates:
# if pTemplate.isInteresting() > 0:
# # term is informative, keep it
# aPopulationTemplates.append(pTemplate)
# stats['population'] = self.countMatches(aPopulationTemplates, \
# self.populationTemplates, errorOut)
return stats
