def test():
authorities = get_authorities()
algo_ii.populate_iks_dict()
report = open( 'report.txt', 'wb' )
data = open('papers30000.csv', 'r')
reader = csv.reader(data, delimiter=',', quotechar='\"')
total_hits = 0
count = 0
total_citations = 0
for row in reader:
index = row[1]
count += 1
citations = row[6].split(';')
keywords = row[9].split(';')
del keywords[-1]
total_citations += len(citations)
recommendations = algo_ii.predict_citations( keywords, authorities )
hits = 0
for citation in citations:
if recommendations.has_key(citation):
hits += 1
total_hits += hits
print( index + " accuracy: " + str(hits) + "/" + str(len(citations)) + "\n" )
report.write(index + " accuracy: " + str(hits) + "/" + str(len(citations)) + "\n")
if (count % 50) == 0:
print( "temp is:" )
print( total_hits )
print( "out of " )
print( total_citations )
print( "achieved accuracy:" )
print( total_hits/float(total_citations) )
report.close()
data.close()
def test_models( FULL_SIM, models_files ):
test_papers = pd.read_csv( TEST_FILEPATH )
# NOTE: Only need for testing with AII:
keywords_docsrels = populate_iks_dict()
authorities = initialize_authorities()
for mod_f in models_files:
print( 'Testing '+ mod_f )
model = Doc2Vec.load( mod_f )
print( 'Model loaded.' )
test_model( FULL_SIM, model, test_papers, keywords_docsrels, authorities )
def specific_test():
import pandas as pd
# NOTE: The first row of data-kw.csv should be the header, i.e.:
# "INDEX","TITLE","AUTHORS","YEAR","PUB_VENUE","REF_ID","REF_NUM","ABSTRACT","KEYWORDS"
papers = pd.read_csv( 'data-kw.csv' )
authorities = get_authorities()
algo_ii.populate_iks_dict()
print( "HUMAN EVAL #1:\n" )
_specific_test( papers, authorities, ['105542', '586892', '695628', '209104', '139162'] )
print( "\n=================================\n" )
print( "HUMAN EVAL #2:\n" )
_specific_test( papers, authorities, ['751328', '619377', '686318', '283022', '591411'] )
print( "\n=================================\n" )
print( "HUMAN EVAL #3:\n" )
_specific_test( papers, authorities, ['360556', '1022648', '1071218', '1112586', '451082'] )
print( "\n=================================\n" )
print( "HUMAN EVAL #4:\n" )
_specific_test( papers, authorities, ['90992', '784131', '1080100', '96640', '503999'] )
print( "\n=================================\n" )
title = list( papers[papers['INDEX'] == pid]['TITLE'] )[0]
authors = list( papers[papers['INDEX'] == pid]['AUTHORS'] )[0]
summary = list( papers[papers['INDEX'] == pid]['ABSTRACT'] )[0]
year = list( papers[papers['INDEX'] == pid]['YEAR'] )[0]
jso['title'] = title
jso['authors'] = authors
jso['summary'] = summary.replace('\n', ' ')
jso['year'] = str(year)
recs.append( jso )
except: # Probably an IndexError (which happens when 'pid' not in DB), but cover our bases
pass
#return json.dumps( recs )
return recs # NOTE: Not actually JSON; it's a list of dicts
def load_papers():
import pandas as pd
return pd.read_csv( 'data-kw.csv' )
if __name__ == '__main__':
if 'tagmap1' in DOC2VEC_MODEL_FILE:
USE_DBLP_IDS = False
else:
USE_DBLP_IDS = True
d2v_model = initialize_d2v_model()
keywords_docsrels = populate_iks_dict()
authorities = initialize_authorities()
papers = load_papers()
serve( d2v_model, keywords_docsrels, authorities, papers )
