def read_rdf(ont1, ont2):
largebio_data_processed_path = 'data/df_largebio_{}_{}.csv'.format(
ont1, ont2)
largebio_ref_processed_path = 'data/df_largebio_{}_{}_ref.csv'.format(
ont1, ont2)
if not os.path.isfile(largebio_data_processed_path):
# Specify path for the alignments and reference alignments
res_dir = os.path.join("data", "largebio-results-2019")
ref_path = os.path.join(
"data", "oaei2019_umls_flagged_reference",
"oaei_{}_{}_mappings_with_flagged_repairs.rdf".format(ont1, ont2))
# Load rdf data
df_data, df_ref = u.load_rdf('largebio', res_dir, ref_path, ont1, ont2)
# Negative sampling
df_data = u.negative_sampling_target(lb_measures, df_data, df_ref)
# Save results to csv
largebio_data_processed_path = 'data/df_largebio_{}_{}.csv'.format(
ont1, ont2)
largebio_ref_processed_path = 'data/df_largebio_{}_{}_ref.csv'.format(
ont1, ont2)
# Save results to csv
df_data.to_csv(largebio_data_processed_path, index=False)
df_ref.to_csv(largebio_ref_processed_path, index=False)
else:
print('File already exists')
df_data = pd.read_csv(largebio_data_processed_path)
df_ref = pd.read_csv(largebio_ref_processed_path)
return df_data, df_ref
'measure_logmap', 'measure_logmaplt', 'measure_ontmat1', 'measure_sanom',
'measure_wiktionary'
]
conference_data_processed_path = 'data/df_conference.csv'
res_dir = os.path.join('data', 'conference-data')
if not os.path.isfile(conference_data_processed_path):
dfs_data, dfs_refs = [], []
for ont1, ont2 in itertools.combinations(cf_ontologies, 2):
ref_path = os.path.join(
"data",
"conference-ref-data",
"{}-{}.rdf".format(ont1, ont2),
)
df_data, df_ref = u.load_rdf('conference', res_dir, ref_path, ont1,
ont2)
df_data = u.negative_sampling_target(cf_measures, df_data, df_ref)
df_data["ontologies"] = f"{ont1}-{ont2}"
dfs_data.append(df_data)
dfs_refs.append(df_ref)
df_conf = pd.concat(dfs_data, ignore_index=True)
df_conf.to_csv(conference_data_processed_path, index=False)
else:
df_conf = pd.read_csv(conference_data_processed_path)
X_cf, y_cf = df_conf[cf_measures], df_conf['label']
#fill missing values with 0
X_cf = X_cf.fillna(0)
#binary features