def __init__(self, name='Meta-DecisionTree', commands=[]):
""" Initializes the TransitivityInt model.
Parameters
----------
name : str
Unique name of the model. Will be used throughout the ORCA
framework as a means for identifying the model.
"""
SentimentAnalyzer.initialize()
self.parameter = {}
self.componentKeys = Keys.person + Keys.task
super().__init__(name, ['misinformation'], ['single-choice'], commands)
def __init__(self, name='Fast-Frugal-Tree', commands = []):
""" Initializes the TransitivityInt model.
Parameters
----------
name : str
Unique name of the model. Will be used throughout the ORCA
framework as a means for identifying the model.
"""
SentimentAnalyzer.initialize()
self.parameter = {}
#self.parameter['thresh'] = 10
self.componentKeys = ['Exciting_Party_Combined', 'Familiarity_Party_Combined', 'Partisanship_Party_Combined','Worrying_Party_Combined','Importance_Party_Combined', 'Partisanship_All_Combined', 'Partisanship_All_Partisan']
super().__init__(name, ['misinformation'], ['single-choice'], commands)
def __init__(self, name='Fast-Frugal-Tree-ZigZag(Z+)', commands=[]):
""" Initializes the model.
Parameters
----------
name : str
Unique name of the model. Will be used throughout the ORCA
framework as a means for identifying the model.
"""
SentimentAnalyzer.initialize()
self.parameter = {}
#self.parameter['thresh'] = 1
self.fft = None
self.lastnode = None
self.componentKeys = Keys.person + Keys.task
super().__init__(name, ['misinformation'], ['single-choice'], commands)
def toDFFormat(self, trialList):
featList = []
for item in trialList:
if item.conservatism < 3.5:
componentKeys = [a.replace('Party','Democrats') for a in self.componentKeys]
if item.conservatism > 3.5:
componentKeys = [a.replace('Party','Republicans') for a in self.componentKeys]
newPars = [item.birep, item.conservatism, item.gender, #int(item.truthful),
item.crt, item.cons, item.accimp]
newPars += [item.itemComponents[a] for a in componentKeys] + [SentimentAnalyzer.analysis(item)[a] for a in SentimentAnalyzer.relevant]
newPars += [item.panas[a] for a in item.panas.keys()] + [item.media[a] for a in item.media.keys()]
featList.append(newPars)
compKeys = [a[:3] + a.split('_')[1][:1] + a.split('_')[2][:1] for a in self.componentKeys]
media = ['Media1_Some people t','Media1_In presenting','Media3_1_To what ext','Media3_2_To what ext','Media3_3_To what ext','Media3_11_To what ex','Media3_12_To what ex']
panas = ['PANAS_' + str(i) for i in range(1, 21)]
cat_columns = ['ct','Gender'] #'truthful']
nr_columns = ['crt', 'cons','AccImp'] + compKeys + SentimentAnalyzer.relevant + panas + media
data_columns = ['response'] + cat_columns + nr_columns
data = pd.DataFrame(data=featList, columns=data_columns)
for col in cat_columns:
data[col] = data[col].astype('category')
for col in nr_columns:
if data[col].dtype != 'float' and data[col].dtype != 'int':
print('type error: ' + data[col])
data.loc[data[col] == '?', col] = np.nan
data[col] = data[col].astype('float')
data['response'] = data['response'].apply(lambda x: True if x==1 else False).astype(bool)
return data
def predictS(self, trial):
analysis = SentimentAnalyzer.analysis(trial)
p = 0
for a in self.relevant:
p += analysis[a] * self.parameter[a]
#print(p)
return 1 if self.thresh < p else 0
def itemsList(source):
itemsList = []
linecount = 0
indcount = 0
line1 = True
lines = open('/home/hippo/git/news-reasoning/benchmarks/misinformation/' +
source + '.csv')
ind = {}
for line in lines:
listLine = line.replace('\r', '').replace('\n', '').split(',')
if line1:
line1 = False
for key in listLine:
ind[key] = indcount
indcount += 1
continue
panasPos = [
'PANAS_' + str(i) for i in [1, 3, 5, 9, 10, 14, 16, 17, 19]
]
panasNeg = [
'PANAS_' + str(i) for i in [2, 4, 6, 7, 8, 11, 12, 13, 15, 18, 20]
]
personfeatures = {
'crt': float(listLine[ind['crt']]),
'conservatism': listLine[ind['conservatism']],
'ct': int(listLine[ind['ct']]),
'education': int(listLine[ind['education']]),
'gender': listLine[ind['gender']],
'accimp': float(listLine[ind['accimp']]),
'panasPos': float(listLine[ind['panasPos']]),
'panasNeg': float(listLine[ind['panasNeg']]),
'reaction_time': float(listLine[ind['reaction_time']])
}
taskfeatures = {}
os.chdir(
"/home/hippo/git/news-reasoning/benchmarks/misinformation/data/study2"
)
for modelfile in glob.glob("with*"):
no = False
for a in ['ecision', 'Pers', 'inear', 'ecommend', 'Tree']:
if a in modelfile:
no = True
if no:
continue
taskfeatures[modelfile] = float(listLine[ind[modelfile]])
if Keys.person == None:
Keys.person = [a for a in personfeatures.keys()]
if Keys.task == None:
SentimentAnalyzer.initialize()
Keys.task = [
a for a in taskfeatures.keys() if 'ikelihood' not in a
]
Keys.task += [a for a in SentimentAnalyzer.relevant]
linecount += 1
itemsList.append(
Item(listLine[ind['id']],
listLine[ind['domain']],
listLine[ind['task']],
listLine[ind['response_type']],
listLine[ind['choices']],
listLine[ind['trial']],
float(listLine[ind['crt']]),
bool('T' in listLine[ind['realnews']]),
bool('T' in listLine[ind['acc']]),
bool('T' in listLine[ind['birep']]),
str(listLine[ind['response']]),
persFeat=personfeatures,
taskFeat=taskfeatures))
return itemsList
def itemsList(source):
itemsList = []
linecount = 0
indcount = 0
line1 = True
lines = open(
'/home/hippo/git/news-reasoning/benchmarks/misinformation/data/' +
source + '.csv')
ind = {}
for line in lines:
listLine = line.replace('\r', '').replace('\n', '').split(',')
if line1:
line1 = False
for key in listLine:
ind[key] = indcount
indcount += 1
continue
linecount += 1
if listLine[ind['_accurate']] == '':
continue
crtresults = []
itemComponents = {}
panas = {}
media = {}
if 'DemRep_C_current' in ind.keys():
if '' == listLine[ind['DemRep_C_current']]:
continue
conservatism = float(listLine[ind['DemRep_C_current']])
elif 'Conservatism' in ind.keys():
conservatism = float(listLine[ind['Conservatism']]) * 7 / 5
elif 'Conserv' in ind.keys():
try:
conservatism = float(listLine[ind['Conserv']]) * 7 / 5
except:
continue
if '' == conservatism:
continue
for item in ind.keys():
if item in [
'Exciting_Democrats_Combined',
'Exciting_Republicans_Combined',
'Familiarity_Democrats_Combined',
'Familiarity_Republicans_Combined',
'Importance_Democrats_Combined',
'Importance_Republicans_Combined',
'Likelihood_Democrats_Combined',
'Likelihood_Republicans_Combined',
'Partisanship_All_Combined', 'Partisanship_All_Partisan',
'Partisanship_Democrats_Combined',
'Partisanship_Republicans_Combined',
'Worrying_Democrats_Combined',
'Worrying_Republicans_Combined'
]: #'Sharing_Democrats_Combined', 'Sharing_Republicans_Combined',
itemComponents[item] = float(listLine[ind[item]])
elif 'CRT1' in item or 'CRT3' in item:
crtresults.append(
float(listLine[ind[item]] in item.split('_')[1].split(':')
[1:]))
elif 'PANAS' in item and 'Inst' not in item:
if listLine[ind[item]] == '':
#logstr('One panas error', listLine[ind['id']], item)
panas[item.split('_T')[0]] = None
continue
panas[item.split('_T')[0]] = float(listLine[ind[item]])
elif 'AccImp_How important' in item:
accimp = float(listLine[ind[item]])
elif 'Gender_What' in item:
gender = float(listLine[ind[item]])
elif 'Sex' == item:
try:
gender = float(listLine[ind[item]])
except:
gender = 'NoInfo'
crt = sum(crtresults) / len(crtresults)
panasPos = [
'PANAS_' + str(i) for i in [1, 3, 5, 9, 10, 14, 16, 17, 19]
]
panasNeg = [
'PANAS_' + str(i) for i in [2, 4, 6, 7, 8, 11, 12, 13, 15, 18, 20]
]
for a in ['_RT_2_Timing-Last Click', '_RT_accurate', '_A_RT_accurate']:
if a not in ind.keys():
continue
try:
react_time = float(listLine[ind[a]])
except:
react_time = 'NoInfo'
if react_time == 'NoInfo':
continue
for a in ['ClintonTrump', 'POTUS2016_Who did yo']:
if a not in ind.keys():
continue
try:
ct = int(listLine[ind[a]])
except:
ct = 'NoInfo'
print('ct:', listLine[ind[a]])
for a in ['Education', 'Education_What is th']:
if a not in ind.keys():
continue
try:
edu = int(listLine[ind[a]])
except:
edu = 'NoInfo'
if 'AccImp_How important' not in ind.keys():
accimp = 'NoInfo'
personfeatures = {
'crt':
float(crt),
'conservatism':
conservatism,
'ct':
ct,
'education':
edu,
'gender':
gender,
'accimp':
accimp,
'panasPos':
sum(panas[a] for a in panasPos
if panas[a] != None) if len(panas.keys()) > 0 else 'NoInfo',
'panasNeg':
sum(panas[a] for a in panasNeg
if panas[a] != None) if len(panas.keys()) > 0 else 'NoInfo',
'reaction_time':
react_time
}
nokeys = []
# for a in personfeatures.keys():
# if personfeatures[a] == 'NoInfo':
# nokeys.append(a)
# for a in nokeys:
# personfeatures.pop(a)
taskfeatures = {}
for a in itemComponents.keys():
if conservatism >= 3.5:
if 'Democrats' in a:
continue
else:
key = a.replace('Republicans', 'Party')
elif conservatism <= 3.5:
if 'Republicans' in a:
continue
else:
key = a.replace('Democrats', 'Party')
taskfeatures[key] = itemComponents[a]
if Keys.person == None:
Keys.person = [a for a in personfeatures.keys()]
if Keys.task == None:
SentimentAnalyzer.initialize()
Keys.task = [
a for a in taskfeatures.keys() if 'ikelihood' not in a
]
Keys.task += [a for a in SentimentAnalyzer.relevant]
newItem = Item(listLine[ind['id']],
listLine[ind['domain']],
listLine[ind['task']],
listLine[ind['response_type']],
listLine[ind['choices']],
listLine[ind['sequence']],
float(crt),
bool(int(listLine[ind['truthful']])),
bool(int(listLine[ind['_accurate']])),
bool(int(listLine[ind['binaryResponse']])),
str(listLine[ind['response']]),
persFeat=personfeatures,
taskFeat=taskfeatures)
itemsList.append(newItem)
return itemsList