# Helper functions
adder = lambda x, y: x + y
def normalize_phrase(phrase):
if type(phrase) == str:
return phrase.lower()
else:
if np.isnan(phrase):
return ''
else:
assert False
argparser.add_argument('-b',
type=str,
help='Target building name',
dest='building')
args = argparser.parse_args()
building = args.building
# Transform parsed lables used in Scrabble for Oracle.
df = pd.read_csv('{0}_rawmetadata.csv'.format(building))\
.set_index('SourceIdentifier')
with open('{0}_char_label_dict.json'.format(building), 'r') as fp:
char_label_dict = json.load(fp)
parse_dict = dict()
from jasonhelper import argparser
argparser.add_argument('-bl', type='slist', dest='buildings')
argparser.add_argument('-nl', type='ilist', dest='sample_nums')
argparser.add_argument('-t', type=str, dest='target_building')
argparser.add_argument('-step', type=int, dest='step_num')
argparser.add_argument('-iter', type=int, dest='iter_num')
sentence = ''.join(
[row[0] for row in fullparsing['VendorGivenName']])
rawpoint.metadata = {'VendorGivenName': sentence}
rawpoint.save()
print('Finished adding full parsing')
# add tagsets
def parse_tagsets(building):
with open('../groundtruth/{0}_tagsets.json'.format(building), 'r') as fp:
true_tagsets = json.load(fp)
for srcid, tagsets in true_tagsets.items():
point = LabeledMetadata.objects(srcid=srcid, building=building)\
.upsert_one(srcid=srcid, building=building)
point.tagsets = tagsets
point_tagset = sel_point_tagset(tagsets)
if not point_tagset:
point_tagset = 'none'
point.point_tagset = point_tagset
point.save()
if __name__ == '__main__':
argparser.add_argument('-b', type=str, dest='building', required=True)
args = argparser.parse_args()
building = args.building
if building in CMU_BUILDINGS:
parse_tagsets(building)
parse_fullparsing(building, write_rawmetadata=True)
def remove_invalid_srcids(building):
with open('config/invalid_srcids.json', 'r') as fp:
invalid_srcids_dict = json.load(fp)
if building not in invalid_srcids_dict:
return None
invalid_srcids = invalid_srcids_dict[building]
print('{0} invalid points are removed.'.format(len(invalid_srcids)))
for srcid in invalid_srcids:
RawMetadata.objects(srcid=srcid).delete()
LabeledMetadata.objects(srcid=srcid).delete()
if __name__ == '__main__':
argparser.add_argument('-b', type=str, dest='building', required=True)
argparser.add_argument('-top',
type='bool',
dest='topclass_flag',
default=False)
# add raw metadata
args = argparser.parse_args()
building = args.building
if building == 'uva_cse':
load_uva_building(building)
print('UVA CSE Done')
sys.exit()
elif building in UCB_BUILDINGS:
extract_raw_ucb_labels()
basedir = './groundtruth/'
if max_score < 0.75:
return None
# TODO: Selection mechanism needs to be more smart.
# Currently mapping
# {'pressure', 'filter', 'discharge', 'his', 'weatherPoint', 'air', 'sp'}
# ->
# "Discharge_Air"
cand = ''
for tagset, score in score_dict.items():
if score == max_score:
if len(cand.split('_')) < len(tagset.split('_')):
cand = tagset
return cand
argparser.add_argument('-src', type=str, dest='srcfile',
help='Haystack instance in JSON or Zinc')
args = argparser.parse_args()
################ Init base information ##############
# Load Brick
brick_tag_g = rdflib.Graph()
brick_tag_g.parse('brick/BrickTag.ttl', format='turtle')
res = brick_tag_g.query("""
select ?s where {
?s <http://www.w3.org/2000/01/rdf-schema#subClassOf>+ <https://brickschema.org/schema/1.0.1/BrickFrame#Tag> .
}""")
brick_tags = [row[0].split('#')[-1] for row in res]
with open('brick/hb_map.json', 'r') as fp:
hb_map = json.load(fp)