def list_workspaces(iam_apikey, version, url):
authenticator = IAMAuthenticator(iam_apikey)
if WATSON_SERVICE != 'nlc':
c = AssistantV1(version=version, authenticator=authenticator)
c.set_service_url(url)
return c.list_workspaces()
else:
c = NaturalLanguageClassifierV1(authenticator)
c.set_service_url(url)
return c.list_classifiers()
def createClassifier():
authenticator = IAMAuthenticator('API_KEY')
natural_language_classifier = NaturalLanguageClassifierV1(authenticator=authenticator)
natural_language_classifier.set_service_url('SERVICE_URL')
with open('./labels.csv', 'rb') as training_data:
classifier = natural_language_classifier.create_classifier(
training_data=training_data,
training_metadata='{"name": "Classifier","language": "en"}'
).get_result()
print(json.dumps(classifier, indent=2))
def checkWholeDoc(fulltext):
#Authenticate API
authenticator = IAMAuthenticator('API_KEY')
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url('SERVICE_URL')
# Process whole doc
for i in fulltext:
classes = natural_language_classifier.classify('35c0a4x769-nlc-127',
i).get_result()
topClass = classes["top_class"]
if topClass in FLAGGED_CLASSES:
return True
def main():
output = {}
# Load the classifier
authenticator = IAMAuthenticator(
'd61RIxMA4RwhewIoThcevX0xJqAo80mMyAnkjwNb8ePy')
discovery = DiscoveryV1(version='2018-08-01', authenticator=authenticator)
discovery.set_service_url(
'https://gateway.watsonplatform.net/natural-language-classifier/api')
classifier = NLClassifier(authenticator)
# Take Input, can input multiple lines
with open('../data/test_data.csv', newline='') as csvfile:
reader = csv.reader(csvfile)
for row in reader:
text = row[0]
# Clean the text
for mark in del_marks:
text = text.replace(mark, '')
text = text.replace(''', '\'')
# Classify
result = classifier.classify('90dbdex665-nlc-266',
text).get_result()
confidence = result['classes'][0]['confidence']
normalized_confidence = (confidence - min_confidence) / (
max_confidence - min_confidence)
output[text] = normalized_confidence
csvfile.close()
# Write output
file = open('output.txt', 'w+')
for out in output:
string = out + ', ' + str(output[out])
file.write(string)
file.close()
def delete_workspaces(iam_apikey, url, version, workspace_ids):
""" Delete workspaces
"""
authenticator = IAMAuthenticator(iam_apikey)
for workspace_id in workspace_ids:
if 'natural-language-classifier' in url:
c = NaturalLanguageClassifierV1(authenticator=authenticator)
c.set_service_url(url)
c.delete_classifier(classifier_id=workspace_id)
else:
c = AssistantV1(version=version, authenticator=authenticator)
c.set_service_url(url)
c.delete_workspace(workspace_id=workspace_id)
print('Cleaned up workspaces')
def processAddresses(lines, lineNums):
#Authenticate API
authenticator = IAMAuthenticator('API_KEY')
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator
)
natural_language_classifier.set_service_url('SERVICE_URL')
boxes = []
for l in lineNums:
for word in lines[l]["words"]:
classes = natural_language_classifier.classify('35c0a4x769-nlc-127',word["text"]).get_result()
topClass = classes["top_class"]
if topClass == "address":
boxes.append(word["boundingBox"][0:2] + word["boundingBox"][4:7])
boxes.append(word["boundingBox"][2:4] + word["boundingBox"][6:8])
return boxes
def training(f_name, api_key, model_name):
authenticator = IAMAuthenticator(api_key)
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url(
'https://gateway.watsonplatform.net/natural-language-classifier/api')
with open(os.path.join(os.path.dirname('__file__'), f_name),
'rb') as training_data:
metadata = json.dumps({'name': model_name, 'language': 'en'})
classifier = natural_language_classifier.create_classifier(
training_metadata=metadata,
training_data=training_data).get_result()
subprocess.call(f'rm {f_name}', shell=True)
return classifier
def processLines(lines):
#Authenticate API
authenticator = IAMAuthenticator('API_KEY')
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url('SERVICE_URL')
# Store lines that have a term
labels = {"race": [], "color": [], "address": []}
# Check each line in the document
for i in range(len(lines)):
classes = natural_language_classifier.classify(
'35c0a4x769-nlc-127', lines[i]["text"]).get_result()
topClass = classes["top_class"]
if topClass in FLAGGED_CLASSES:
# print(topClass, ":", lines[i]["text"])
labels[topClass].append(i)
return labels
def delete_workspaces(iam_apikey, url, version, workspace_ids, auth_type):
""" Delete workspaces
"""
if auth_type == 'iam':
authenticator = IAMAuthenticator(iam_apikey)
elif auth_type == 'bearer':
authenticator = BearerTokenAuthenticator(iam_apikey)
else:
raise ValueError(f'Unknown auth_type "{auth_type}"')
for workspace_id in workspace_ids:
if 'natural-language-classifier' in url:
c = NaturalLanguageClassifierV1(authenticator=authenticator)
c.set_service_url(url)
c.delete_classifier(classifier_id=workspace_id)
else:
c = AssistantV1(version=version, authenticator=authenticator)
c.set_service_url(url)
c.delete_workspace(workspace_id=workspace_id)
print('Cleaned up workspaces')
def delete():
model_name = request.json['model_name']
api_key = request.json['api_key']
authenticator = IAMAuthenticator(api_key)
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url(
'https://gateway.watsonplatform.net/natural-language-classifier/api')
classifier_id = None
for c in natural_language_classifier.list_classifiers().get_result(
)['classifiers']:
if c['name'] == model_name:
classifier_id = c['classifier_id']
if not classifier_id:
return {'message': "Classifier not found, check its name again"}
else:
natural_language_classifier.delete_classifier(classifier_id)
res = {'message': f"Model '{model_name}' Deleted"}
return jsonify(res)
def func(args):
classifier_name = ''
classifier_description = ''
classes = []
language = 'en'
counterexamples = []
metadata = {}
learning_opt_out = False
authenticator = choose_auth(args)
nlc = NaturalLanguageClassifierV1(authenticator=authenticator)
nlc.set_service_url(args.url)
classifier_name = "My Classifier"
if args.classifier_name is not None:
classifier_name = args.classifier_name
metadata = {"name": classifier_name, "language": "en"}
if args.trainingFile is not None:
with open(args.trainingFile, 'rb') as training_data:
classifier = nlc.create_classifier(
training_data=training_data,
training_metadata=json.dumps(metadata)).get_result()
resp = classifier
# Poke the training status every SLEEP_INCRE secs
sleep_counter = 0
while sleep_counter < TIME_TO_WAIT:
raw_resp = nlc.get_classifier(classifier_id=resp[CLASSIFIER_ID_TAG])
resp = raw_resp.get_result()
if resp['status'] == 'Available':
print(json.dumps(resp, indent=4)) # double quoted valid JSON
return
sleep_counter += SLEEP_INCRE
sleep(SLEEP_INCRE)
raise TrainTimeoutException('NLC training timeout')
import json
from ibm_watson import NaturalLanguageClassifierV1
service = NaturalLanguageClassifierV1(
url='https://gateway.watsonplatform.net/natural-language-classifier/api',
iam_apikey='JieYHJwBRgrd5Rl9R4q63d5DWvAuffdrRIj1jKhkfoAH')
classes = service.classify('8a423bx518-nlc-1830',
'Airline lost my luggage twice in the same month, delayed my flight, ' + \
'and didn\'t care about my stopover!').get_result()
print(json.dumps(classes, indent=2))
def func(args):
in_df = None
out_df = None
test_column = UTTERANCE_COLUMN
if args.test_column is not None: # Test input has multiple columns
test_column = args.test_column
in_df = pd.read_csv(args.infile,
quoting=csv.QUOTE_ALL,
encoding=UTF_8,
keep_default_na=False)
if test_column not in in_df: # Look for target test_column
raise ValueError(
"Test column {} doesn't exist in file.".format(test_column))
if args.merge_input: # Merge rest of columns from input to output
out_df = in_df
else:
out_df = in_df[[test_column]].copy()
out_df.columns = [test_column]
else:
test_series = pd.read_csv(args.infile,
quoting=csv.QUOTE_ALL,
encoding=UTF_8,
header=None,
squeeze=True,
keep_default_na=False)
if isinstance(test_series, pd.DataFrame):
raise ValueError('Unknown test column')
# Test input has only one column and no header
out_df = test_series.to_frame()
out_df.columns = [test_column]
# Initial columns for test output
for column in test_out_header:
out_df[column] = ''
# Applied coroutines
sem = asyncio.Semaphore(args.rate_limit)
loop = asyncio.get_event_loop()
authenticator = choose_auth(args)
nlc = NaturalLanguageClassifierV1(authenticator=authenticator)
nlc.set_service_url(args.url)
tasks = (fill_df(out_df.loc[row_idx, test_column], row_idx, out_df,
args.workspace_id, nlc, sem)
for row_idx in range(out_df.shape[0]))
loop.run_until_complete(asyncio.gather(*tasks))
loop.close()
if args.golden_intent_column is not None:
golden_intent_column = args.golden_intent_column
if golden_intent_column not in in_df.columns:
print("No golden intent column '{}' is found in input.".format(
golden_intent_column))
else: # Add INTENT_JUDGE_COLUMN based on golden_intent_column
out_df[INTENT_JUDGE_COLUMN] = \
(in_df[golden_intent_column]
== out_df[PREDICTED_INTENT_COLUMN]).map(BOOL_MAP)
out_df[SCORE_COLUMN] = \
out_df[INTENT_JUDGE_COLUMN].map({'yes': 1, 'no': 0})
if args.partial_credit_table is not None:
credit_tables = parse_partial_credit_table(args.partial_credit_table)
for row_idx in range(out_df.shape[0]):
golden_intent = out_df.loc[row_idx,
args.golden_intent_column].strip()
predict_intent = out_df.loc[row_idx,
PREDICTED_INTENT_COLUMN].strip()
if golden_intent == predict_intent:
out_df.loc[row_idx, SCORE_COLUMN] = 1.0
elif golden_intent not in credit_tables or \
predict_intent not in credit_tables[golden_intent]:
out_df.loc[row_idx, SCORE_COLUMN] = 0
else:
out_df.loc[row_idx, SCORE_COLUMN] = \
credit_tables[golden_intent][predict_intent]
save_dataframe_as_csv(df=out_df, file=args.outfile)
print("Wrote test result file to {}".format(args.outfile))
def deleteClassifier():
authenticator = IAMAuthenticator('API_KEY')
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url('SERVICE_URL')
natural_language_classifier.delete_classifier('35c0a4x769-nlc-127')
import json
import os
import requests
from dotenv import load_dotenv
from flask import Flask, render_template, request
from ibm_watson import NaturalLanguageClassifierV1
DEBUG = True
app = Flask(__name__)
load_dotenv(os.path.join(os.path.dirname(__file__), ".env"))
classifier_id = os.environ.get("CLASSIFIER_ID")
NLC_SERVICE = NaturalLanguageClassifierV1()
@app.route('/')
def default():
classifier_info = "cannot detect classifier"
if NLC_SERVICE:
classifier_info = ("classifier detected, using API: " +
NLC_SERVICE.service_url)
return render_template('index.html',
classifier_info=classifier_info,
icd_code="",
icd_output="",
classifier_output="")
def testing(df_test, model_name, api_key, email_id=[], emailed=False):
authenticator = IAMAuthenticator(api_key)
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url('https://gateway.watsonplatform.net/natural-language-classifier/api')
classifier_id = None
for c in natural_language_classifier.list_classifiers().get_result()['classifiers']:
if c['name']==model_name:
classifier_id = c['classifier_id']
if not classifier_id:
return {'message':"Classifier not found, check its name again"}
df_test = pd.DataFrame(df_test)
# print(type(df_test), df_test.shape, df_test.columns)
df_test.index = df_test.index.astype(int)
df_test['text'] = df_test['text'].str[:1024]
df_test['text'] = df_test['text'].apply(lambda x: ((x.encode("unicode_escape").decode("utf-8"))[:1024]).strip())
y_true = df_test['label'].to_numpy()
y_pred = []
conf = []
numbers = []
while(True):
status = natural_language_classifier.get_classifier(classifier_id).get_result()['status']
if (status=='Training') and (emailed==False):
return {'message':"Classifier not Trained, try when its available"}
if status=='Training':
continue
for num, example in zip(df_test['id'], df_test['text']):
numbers.append(num)
classes = natural_language_classifier.classify(classifier_id, example).get_result()
# print(classes)
pred_label = classes['top_class']
pred_conf = classes['classes'][0]['confidence']
# print(pred_label, pred_conf)
y_pred.append(pred_label)
conf.append(pred_conf)
break
y_pred = np.array(y_pred, dtype=np.object)
numbers = np.array(numbers, dtype=np.object)
conf = np.array(conf, dtype=np.float64)
df_result = pd.DataFrame(columns=['real','pred','conf','id'])
df_result['id'] = numbers
df_result['pred'] = y_pred
df_result['conf'] = conf
df_result['real'] = y_true
report = classification_report(y_true, y_pred, output_dict=True)
report = pd.DataFrame(report).T
# report = report[~report.index.isin(['accuracy', 'macro avg', 'weighted avg'])]
report_txt = classification_report(y_true, y_pred)
conf_mat = pd.crosstab(df_result['real'], df_result['pred'], rownames=['Actual'], colnames=['Pred'])
acc = accuracy_score(y_true, y_pred)
global do_ml_flow
if do_ml_flow:
log_metric("Test Accuracy", acc)
df_result.to_excel(f'tmp/df_result -- {model_name}.xlsx', index=False)
report.to_excel(f'tmp/report -- {model_name}.xlsx')
with open(f'tmp/report -- {model_name}.txt', 'w') as f:
print(report_txt, file=f)
conf_mat.to_excel(f'tmp/conf_mat -- {model_name}.xlsx')
log_artifact(f'tmp/df_result -- {model_name}.xlsx')
log_artifact(f'tmp/report -- {model_name}.xlsx')
log_artifact(f'tmp/report -- {model_name}.txt')
log_artifact(f'tmp/conf_mat -- {model_name}.xlsx')
subprocess.call(f'rm "tmp/df_result -- {model_name}.xlsx"', shell=True)
subprocess.call(f'rm "tmp/report -- {model_name}.xlsx"', shell=True)
subprocess.call(f'rm "tmp/report -- {model_name}.txt"', shell=True)
subprocess.call(f'rm "tmp/conf_mat -- {model_name}.xlsx"', shell=True)
if emailed==False:
res = {}
res['df_res'] = df_result.to_dict()
res['acc'] = acc
res['report'] = report.to_dict()
res['report_txt'] = report_txt
res['conf_mat'] = conf_mat.to_dict()
return res
else:
res = {}
res['df_res'] = df_result.to_dict()
res['acc'] = acc
res['report'] = report.to_dict()
res['report_txt'] = report_txt
res['conf_mat'] = conf_mat.to_dict()
with open(f'watson_results/{model_name}.json', 'w') as fp:
json.dump(res, fp)
# natural_language_classifier = NaturalLanguageClassifierV1(
# authenticator=authenticator
# )
# natural_language_classifier.set_service_url('SERVICE_URL')
# status = natural_language_classifier.get_classifier('35ba1fx766-nlc-138').get_result()
# print (json.dumps(status, indent=2))
##################### Run classifier
import json
from ibm_watson import NaturalLanguageClassifierV1
from ibm_cloud_sdk_core.authenticators import IAMAuthenticator
authenticator = IAMAuthenticator('API_KEY')
natural_language_classifier = NaturalLanguageClassifierV1(
authenticator=authenticator)
natural_language_classifier.set_service_url('SERVICE_URL')
classes = natural_language_classifier.classify('35ba1fx766-nlc-138',
'Athens').get_result()
print(json.dumps(classes, indent=2))
##################### Delete classifier
# from ibm_watson import NaturalLanguageClassifierV1
# from ibm_cloud_sdk_core.authenticators import IAMAuthenticator
# authenticator = IAMAuthenticator('API_KEY')
# natural_language_classifier = NaturalLanguageClassifierV1(
# authenticator=authenticator
# )
import json
from ibm_watson import NaturalLanguageClassifierV1
natural_language_classifier = NaturalLanguageClassifierV1(
iam_apikey='{iam-apikey}',
url='https://gateway.watsonplatform.net/natural-language-classifier/api')
classes = natural_language_classifier.classify(
'94904ex626-nlc-66', 'história muito legal sobre a cerveja').get_result()
print(classes['top_class'])
from __future__ import print_function
import json
import os
# from os.path import join, dirname
from ibm_watson import NaturalLanguageClassifierV1
# If service instance provides API key authentication
service = NaturalLanguageClassifierV1(
## url is optional, and defaults to the URL below. Use the correct URL for your region.
url='https://gateway.watsonplatform.net/natural-language-classifier/api',
iam_apikey='YOUR APIKEY')
# service = NaturalLanguageClassifierV1(
# ## url is optional, and defaults to the URL below. Use the correct URL for your region.
# # url='https://gateway.watsonplatform.net/natural-language-classifier/api',
# username='******',
# password='******')
classifiers = service.list_classifiers().get_result()
print(json.dumps(classifiers, indent=2))
# create a classifier
with open(
os.path.join(os.path.dirname(__file__),
'../resources/weather_data_train.csv'),
'rb') as training_data:
metadata = json.dumps({'name': 'my-classifier', 'language': 'en'})
classifier = service.create_classifier(
metadata=metadata, training_data=training_data).get_result()
classifier_id = classifier['classifier_id']
from flask import Flask, render_template, request
from ibm_watson import NaturalLanguageClassifierV1
DEBUG = True
app = Flask(__name__)
load_dotenv(os.path.join(os.path.dirname(__file__), ".env"))
nlc_username = os.environ.get("NATURAL_LANGUAGE_CLASSIFIER_USERNAME")
nlc_password = os.environ.get("NATURAL_LANGUAGE_CLASSIFIER_PASSWORD")
nlc_iam_apikey = os.environ.get("NATURAL_LANGUAGE_CLASSIFIER_IAM_APIKEY")
classifier_id = os.environ.get("CLASSIFIER_ID")
# Use provided credentials from environment or pull from IBM Cloud VCAP
if nlc_iam_apikey != "placeholder":
NLC_SERVICE = NaturalLanguageClassifierV1(iam_apikey=nlc_iam_apikey)
elif nlc_username != "placeholder":
NLC_SERVICE = NaturalLanguageClassifierV1(username=nlc_username,
password=nlc_password)
else:
NLC_SERVICE = NaturalLanguageClassifierV1()
@app.route('/')
def default():
classifier_info = "cannot detect classifier"
if NLC_SERVICE:
classifier_info = "classifier detected, using API: " + NLC_SERVICE.url
return render_template('index.html',
classifier_info=classifier_info,
icd_code="",
import json
import os
import ssl
from ibm_watson import NaturalLanguageClassifierV1
from ibm_cloud_sdk_core.authenticators import IAMAuthenticator
authenticator = IAMAuthenticator(
'yjsnqyHlLV6Hre2gSL9LqtdAaU9hhRBm7Y_f3k8QTS0v')
service = NaturalLanguageClassifierV1(authenticator=authenticator)
service.set_service_url(
'https://api.us-south.assistant.watson.cloud.ibm.com/instances/20b1e02e-12ff-4ab0-87e6-469aa5868021'
)
classifiers = service.list_classifiers().get_result()
print(json.dumps(classifiers, indent=2))
# create a classifier
with open(
os.path.join(os.path.dirname(__file__),
'../resources/weather_data_train.csv'),
'rb') as training_data:
metadata = json.dumps({'name': 'my-classifier', 'language': 'en'})
classifier = service.create_classifier(
training_metadata=metadata, training_data=training_data).get_result()
classifier_id = classifier['classifier_id']
print(json.dumps(classifier, indent=2))
status = service.get_classifier(classifier_id).get_result()
print(json.dumps(status, indent=2))
import json
from ibm_watson import NaturalLanguageClassifierV1
from ibm_watson import ApiException
import Settings
try:
natural_language_classifier = NaturalLanguageClassifierV1(
iam_apikey=Settings.API_KEY, url=Settings.URL)
with open(Settings.TRAINING_DATA_PATH, 'rb') as training_data:
with open(Settings.METADATA_PATH, 'rb') as metadata:
classifier = natural_language_classifier.create_classifier(
training_data=training_data, metadata=metadata).get_result()
print(json.dumps(classifier, indent=2))
except ApiException as ex:
print("Method failed with status code " + str(ex.code) + ": " + ex.message)
'''
Expected response:
{
"name": "TutorialClassifier",
"language": "en",
"status": "Training",
"url": "https://gateway.watsonplatform.net/natural-language-classifier/api/v1/classifiers/0e6935x475-nlc-2948",
"classifier_id": "0e6935x475-nlc-2948",
"created": "2018-12-10T17:42:31.823Z",
"status_description": "The classifier instance is in its training phase, not yet ready to accept classify requests"
}
'''
from __future__ import print_function
import json
import os
# from os.path import join, dirname
from ibm_watson import NaturalLanguageClassifierV1
# If service instance provides API key authentication
service = NaturalLanguageClassifierV1(
## url is optional, and defaults to the URL below. Use the correct URL for your region.
url='https://gateway.watsonplatform.net/natural-language-classifier/api',
iam_apikey='DBxOesEcwYTQK9-dvcaxTwBICWk0s3RwwEW6m-2eppDn')
# service = NaturalLanguageClassifierV1(
# ## url is optional, and defaults to the URL below. Use the correct URL for your region.
# # url='https://gateway.watsonplatform.net/natural-language-classifier/api',
# username='******',
# password='******')
classifiers = service.list_classifiers().get_result()
print(json.dumps(classifiers, indent=2))
# create a classifier
with open(
os.path.join(os.path.dirname(__file__),
'../resources/weather_data_train.csv'),
'rb') as training_data:
metadata = json.dumps({'name': 'my-classifier', 'language': 'en'})
classifier = service.create_classifier(
metadata=metadata, training_data=training_data).get_result()
classifier_id = classifier['classifier_id']
import json
from ibm_watson import NaturalLanguageClassifierV1
natural_language_classifier = NaturalLanguageClassifierV1(
iam_apikey='{apikey}', url='{url}')
status = natural_language_classifier.delete_classifier(
'{classifier_id}').get_result()
print(json.dumps(status, indent=2))
def main():
# Define everything needed later
inputs = []
max_btwness = 0.0
min_btwness = 0.0
avg_btwness = 0.0
# Load the classifier
authenticator = IAMAuthenticator(
'd61RIxMA4RwhewIoThcevX0xJqAo80mMyAnkjwNb8ePy')
discovery = DiscoveryV1(version='2018-08-01', authenticator=authenticator)
discovery.set_service_url(
'https://gateway.watsonplatform.net/natural-language-classifier/api')
classifier = NLClassifier(authenticator)
# Take Input, can input multiple lines
text_file = open('../data/input.txt')
for line in text_file:
line_text = line.split(
',', 2) # maximum 2 splits, because some text contains commas
n1 = -1
n2 = -1
try:
n1 = int(line_text[0])
n2 = int(line_text[1])
except:
print("Node id should be an integer")
# Clean the text
for mark in del_marks:
text = line_text[2].replace(mark, '')
text = text.replace(''', '\'')
# Store this line if input is valid
if not (n1 == -1 or n2 == -1):
input = [n1, n2, text]
inputs.append(input)
text_file.close()
# Get the statistics of the graph
stat_file = open('../data/stat.txt')
for line in stat_file:
if 'max' in line:
max_btwness = float(line.split(': ')[1])
if 'min' in line:
min_btwness = float(line.split(': ')[1])
if 'avg' in line:
avg_btwness = float(line.split(': ')[1])
stat_file.close()
# Get the betweenness record
count = 0
file = open('../data/btwness.txt')
for line in file:
line_text = line.split(',')
n1 = int(line_text[0])
n2 = int(line_text[1])
btwness = float(line_text[2])
# I search if there's an input represents the same edge as appeared in full list
# so that I only need to iterate the full list of edges once,
# because the length of inputs is much smaller than the length of all edges
# and interating through inputs repeatly will be less costly
# than iterating through all edges repeatly
for input in inputs:
# Append btwness as the 4th feature to each input
index = inputs.index(input)
if input[0] == n1:
if input[1] == n2:
input.append(btwness)
count = count + 1 # count how many inputs have the 4th feature
if input[1] == n1:
if input[0] == n1:
input.append(btwness)
count = count + 1
inputs[index] = input
# Don't keep iterating if all inputs are processed
if count == len(inputs):
break
file.close()
for input in inputs:
# Classify
result = classifier.classify('90dbdex665-nlc-266',
input[2]).get_result()
class_name = result['classes'][0]['class_name']
confidence = result['classes'][0]['confidence']
# if top class is 'bully'
if 'bully' in class_name:
normalized_confidence = (confidence - min_confidence) / (
max_confidence - min_confidence)
if len(input) != 4:
# This means their is no betweenness data appended in the previous step
# So the edge does not exist, they are strangers, it is purely bullying
if normalized_confidence > .9:
print(input[2] + "bully")
else:
btwness = input[3]
print(f'{confidence}, {normalized_confidence}, {btwness}')
if normalized_confidence > .9:
if btwness > avg_btwness: # need a better threshold
# larger btwness means less significant network
print(input[2] + "prob. FP")
else:
print(input[2] + "bully")
elif normalized_confidence > .5:
if btwness < avg_btwness: # need a better threshold
# smaller btwness means more significant network
print(input[2] + "prob. FN")
