def upgrade_model_batch():
enable_crossfeature = Config(ALGOCONFIG_PATH).get('enable-cross-feature')
_model = generate_model(REALTIME_MODE, cross_features=enable_crossfeature)
model_name = export_model(_model, REALTIME_MODE, enable_crossfeature)
Config(ALGOCONFIG_PATH).set('model-name/realtime', model_name)
_model = generate_model(BATCH_MODE, cross_features=enable_crossfeature)
model_name = export_model(_model, BATCH_MODE, enable_crossfeature)
Config(ALGOCONFIG_PATH).set('model-name/batch', model_name)
def __init__(self):
access_key_id = Config(DEPLOY_CONFIG_PATH).get('nectar/access_key_id')
access_key_secret = Config(DEPLOY_CONFIG_PATH).get('nectar/access_key_secret')
region = RegionInfo(name='melbourne', endpoint='nova.rc.nectar.org.au')
self.ec2_conn = boto.connect_ec2(aws_access_key_id=access_key_id,
aws_secret_access_key=access_key_secret,
is_secure=True,
region=region,
port=8773,
path='/Services/Cloud',
validate_certs=False)
def export_stats(stats, mode, enable_crossfeature):
json_data = stats.to_json()
stat_name = _generate_stat_name(mode, enable_crossfeature)
filename = MODEL_FILE_BASE_PATH + "{}.json".format(stat_name)
if mode == REALTIME_MODE:
Config(ALGOCONFIG_PATH).set('model-name/train-stats/realtime',
"{}.json".format(stat_name))
else:
Config(ALGOCONFIG_PATH).set('model-name/train-stats/batch',
"{}.json".format(stat_name))
with open(filename, "w") as json_file:
json_file.write(json_data)
def add_instance(self):
instance_type = Config(DEPLOY_CONFIG_PATH).get('nectar/instance_type')
keypair_name = Config(DEPLOY_CONFIG_PATH).get('nectar/keypair_name')
security_groups = Config(DEPLOY_CONFIG_PATH).get('nectar/security_groups')
zone = Config(DEPLOY_CONFIG_PATH).get('nectar/zone')
image_id = self._get_image_id(Config(DEPLOY_CONFIG_PATH).get('nectar/image_name'))
reservation = self.ec2_conn.run_instances(image_id=image_id,
key_name=keypair_name,
instance_type=instance_type,
security_groups=security_groups,
placement=zone)
instance = reservation.instances[0]
instance_info = {"id": instance.id, "ip": instance.private_ip_address, "state": instance.state}
logger.info('Successfully created instance: {}'.format(instance_info))
return instance_info
def _calc(self, data, mode):
model_name = Config(ALGOCONFIG_PATH).get('model-name/{}'.format(
mode_name[mode]))
stat_path = MODEL_FILE_BASE_PATH + Config(ALGOCONFIG_PATH).get(
'model-name/train-stats/{}'.format(mode_name[mode]))
model = import_model(model_name)
train_stats = import_stats(stat_path)
data['vector']['label'] = -1
feature_vectors = generate_feature_vectors(
items=[data], mode=mode, cross_features=enable_crossfeature)
test_set = generate_dataset(data_list=feature_vectors,
mode=mode,
cross_features=enable_crossfeature)
test_set.pop('label')
normed_test_data = norm(test_set, train_stats)
test_predictions = model.predict(normed_test_data).flatten()
return _limit_range(test_predictions[0], lower=0, upper=1)
def __init__(self, db_name):
server_config = Config(CONFIG_PATH).get('couchdb')
self.client = CouchDB(server_config['username'],
server_config['password'],
url=server_config['server_addr'],
connect=True,
auto_renew=True)
self.select_db(db_name)
def uclassify_topics(text):
try:
key = Config(CONFIG_PATH).get('uclassify/apikey')
url = 'https://api.uclassify.com/v1/uClassify/Topics/classify'
data = {'texts': [text]}
header = {'Authorization': 'Token {}'.format(key), 'Content-Type': 'application/json'}
response = requests.post(url=url, data=json.dumps(data), headers=header)
if response.status_code == 200:
resp_data = ast.literal_eval(response.text)[0]['classification']
res = {x['className']: x['p'] for x in resp_data}
return res
else:
keygen = UclassifyKeyGenerator()
new_key = keygen.get_key()
keygen.close()
Config(CONFIG_PATH).set('uclassify/apikey', new_key)
return uclassify_topics(text)
except Exception as ex:
logger.error('Error when uClassifying text: {}'.format(ex))
def add_volume(self, instance_id):
zone = Config(DEPLOY_CONFIG_PATH).get('nectar/zone')
volume = self.ec2_conn.create_volume(size=50, zone=zone)
info = self.get_volume_info(volume.id)
while info["status"] != "available":
logger.info("Volume state: " + info["status"])
time.sleep(3)
info = self.get_volume_info(volume.id)
successful = volume.attach(instance_id, "/dev/vdc")
if successful:
logger.info("volume: " + volume.id + " set and bounded to /dev/vdc")
else:
logger.error("ERROR: volume creating failed.")
class Mrisa:
port = Config(CONFIG_PATH).get('mrisa/port')
service = Config(CONFIG_PATH).get('mrisa/server-path')
proc = None
def start(self):
try:
self.proc = subprocess.Popen(['python3', self.service, '--port', str(self.port)])
time.sleep(3)
logger.info('MRISA service started at port {}.'.format(self.port))
except Exception as ex:
logger.error('Error occured when starting MRISA: {}'.format(ex))
def get_image_info(self, image_url):
data = json.dumps({"image_url": image_url})
url = 'http://localhost/search'
storage = BytesIO()
c = pycurl.Curl()
c.setopt(c.URL, str(url))
c.setopt(c.PORT, self.port)
c.setopt(c.HTTPHEADER, ['Content-Type: application/json'])
c.setopt(pycurl.POST, 1)
c.setopt(pycurl.POSTFIELDS, data)
c.setopt(c.WRITEFUNCTION, storage.write)
c.perform()
c.close()
returned_json = storage.getvalue().decode('UTF-8')
return json.loads(returned_json)
def stop(self):
logger.info('Terminating process pid: {}'.format(self.proc.pid))
self.proc.terminate()
self.proc.kill()
time.sleep(1)
os.system('kill -9 {}'.format(self.proc.pid))
logger.info('MRISA service stopped.')
def __vectorize(self, info1, info2, mode):
# todo: handle modes.
result = {}
profile1, profile2 = info1['profile'], info2['profile']
logger.info('Evaluating usename...')
result['username'] = singleword_similarity(profile1, profile2)
logger.info('Evaluating profile image...')
result['profileImage'] = profile_img_sim(profile1['image'],
profile2['image'])
logger.info('Evaluating self description text...')
result['self_desc'] = self.semantic_sim.similarity(
profile1.get('description', ''), profile2.get('description', ''))
logger.info('Evaluating self description url...')
result['desc_overlap_url_count'] = desc_overlap_url(
{
'platform': info1['platform'],
'username': profile1['username'],
'desc': profile1.get('description', '')
}, {
'platform': info2['platform'],
'username': profile2['username'],
'desc': profile2.get('description', '')
})
posts1 = info1['posts_content'] if 'posts_content' in info1.keys(
) else []
posts2 = info2['posts_content'] if 'posts_content' in info2.keys(
) else []
if mode == REALTIME_MODE and len(posts1) == 0 and len(posts2) == 0:
return result
logger.info('Evaluating writing style...')
result['writing_style'] = writing_style_sim(posts1, posts2)
logger.info('Evaluating post similarity...')
max_post_enabled = bool(
Config(ALGOCONFIG_PATH).get('max-post-similarity-enabled'))
result['post_text'] = self.max_post_sim(
posts1, posts2) if max_post_enabled else self.overall_post_sim(
posts1, posts2)
logger.info('Evaluating uClassify topical similarity...')
result['uclassify'] = uclassify_similarity(
" ".join(_get_post_text(posts1)), " ".join(_get_post_text(posts2)))
logger.info('Calculation finished.')
return result
import ast
import random
import os
from constant import CONFIG_PATH
from similarity.Config import Config
CONFIG = Config(CONFIG_PATH)
PAIRING_FILE_PATH = CONFIG.get("sampler/pairing_file")
MODE = CONFIG.get("sampler/mode")
INSTA_FOLDER = CONFIG.get("sampler/instagram_folder")
TWITTER_FOLDER = CONFIG.get("sampler/twitter_folder")
class Sampler:
def __init__(self):
self.items = []
if MODE == 'file':
with open(PAIRING_FILE_PATH, "r") as file:
while True:
line = file.readline()
if not line:
break
self.items.append(ast.literal_eval(line))
dictlist = [{x['instagram']: x['twitter']} for x in self.items]
self.ins_to_twi_dict = dict(kv for d in dictlist
for kv in d.items())
dictlist = [{x['twitter']: x['instagram']} for x in self.items]
self.twi_to_ins_dict = dict(kv for d in dictlist
for kv in d.items())
def __init__(self):
keyfile = Config(CONFIG_PATH).get('google/keyfile_path')
os.environ["GOOGLE_APPLICATION_CREDENTIALS"] = keyfile
def __init__(self):
subscription_key = Config(CONFIG_PATH).get('microsoft/subscription_key')
self.headers = {'Ocp-Apim-Subscription-Key': subscription_key}
time.sleep(5)
def getResultTable(self):
while True:
try:
target = self.browser.find_element_by_id("ctl00_ContentPlaceHolder1_tblChart")
return target
except Exception as e:
time.sleep(5)
continue
def close(self):
self.browser.quit()
tea_enabled = bool(Config(ALGOCONFIG_PATH).get('tea-enabled'))
def query_writing_style(text):
if len(text) == 0:
return {}
text = ''.join(c for c in text if c <= '\uFFFF')
try:
readbility_metrics = dict(readability.getmeasures(text, lang='en')['readability grades'])
except ValueError:
readbility_metrics = 0.5
logger.warning('Text is Empty, readability return 0.5 as default.')
if tea_enabled:
text = ' '.join(text.split(' ')[:300])
tea = TeaUtils()
tea_metrics = tea.getTextMetrics(text)
def __init__(self):
self.config = Config(CONFIG_PATH)
self.semantic_sim = TensorSimilarity()
import datetime
from constant import CONFIG_PATH
from similarity.Config import Config
INFO = 0
WARNING = 1
ERROR = 2
enable_color = Config(CONFIG_PATH).get('logger/color-enabled')
write_to_file = Config(CONFIG_PATH).get('logger/write-to-file')
log_file_path = Config(CONFIG_PATH).get('logger/logfile-path')
def info(message):
__print_message(message, INFO)
def error(message):
__print_message(message, ERROR)
def warning(message):
__print_message(message, WARNING)
def __print_message(message, level):
color = {INFO: '\033[0m', WARNING: '\033[93m ', ERROR: '\033[91m '}
prefix = {INFO: '(INFO)', WARNING: '(WARNING)', ERROR: '(ERROR)'}
time = str(datetime.datetime.now().strftime("[%Y-%m-%d %H:%M:%S]"))
log_body = '{prefix} {time} {message}'.format(prefix=prefix[level], time=time, message=message)
from automation.batch.ModelUpgrade import import_model, import_stats, generate_dataset, norm, generate_feature_vectors
from constant import ALGOCONFIG_PATH, BATCH_MODE, REALTIME_MODE, MODEL_FILE_BASE_PATH
from similarity.Config import Config
mode_name = {BATCH_MODE: 'batch', REALTIME_MODE: 'realtime'}
enable_crossfeature = Config(ALGOCONFIG_PATH).get('enable-cross-feature')
class OverallSimilarityCalculator:
def _calc(self, data, mode):
model_name = Config(ALGOCONFIG_PATH).get('model-name/{}'.format(
mode_name[mode]))
stat_path = MODEL_FILE_BASE_PATH + Config(ALGOCONFIG_PATH).get(
'model-name/train-stats/{}'.format(mode_name[mode]))
model = import_model(model_name)
train_stats = import_stats(stat_path)
data['vector']['label'] = -1
feature_vectors = generate_feature_vectors(
items=[data], mode=mode, cross_features=enable_crossfeature)
test_set = generate_dataset(data_list=feature_vectors,
mode=mode,
cross_features=enable_crossfeature)
test_set.pop('label')
normed_test_data = norm(test_set, train_stats)
test_predictions = model.predict(normed_test_data).flatten()
return _limit_range(test_predictions[0], lower=0, upper=1)
def calc(self, data):
if has_full_property(data):
return self._calc(data, BATCH_MODE)
return self._calc(data, REALTIME_MODE)
