def __init__(self, ref_file_path):
old_gs = old_read_gs(ref_file_path)
gs = read_gs(ref_file_path)
old_filelist = old_gs.keys()
filelist = gs.keys()
list_check_equality("gold_standard", old_filelist, filelist)
for filename in filelist:
dict_check_equality(filename, old_gs, gs)
def run_experiment(self, crank_flag, single_rank_flag, expand_rank_flag):
ref = read_gs(self.ref_file_path)
eval_score = self.run_evaluation(ref, crank_flag, single_rank_flag, expand_rank_flag)
# print eval_score
for winsize in self.winsize_list:
for cutoff in self.cutoff_list:
for lamb in self.lamb_list:
avg_eval_score = self.get_avg_evaluation(eval_score[winsize][cutoff][lamb])
avg_eval_score['winsize'] = winsize
avg_eval_score['cutoffpath'] = cutoff
avg_eval_score['lambda'] = lamb
avg_eval_score['k1'] = self.k1
avg_eval_score['b'] = self.b
for col_name in self.csv_column:
self.save_fp.write(str(avg_eval_score[col_name]))
if col_name in self.csv_column[:-1]:
self.save_fp.write(',')
elif col_name is self.csv_column[-1]:
self.save_fp.write('\n')
print 'excluded files in evaluation : ', self.excluded_files
def run_experiment(self):
ref = read_gs(self.ref_file_path)
eval_score = self.run_evaluation(ref)
winsize = self.winsize
for k1 in self.k1_list:
for b in self.b_list:
for cutoff in self.cutoff_list:
for lamb in self.lamb_list:
avg_eval_score = self.get_avg_evaluation(eval_score[cutoff][lamb][k1][b])
avg_eval_score['winsize'] = winsize
avg_eval_score['cutoffpath'] = cutoff
avg_eval_score['lambda'] = lamb
avg_eval_score['k1'] = k1
avg_eval_score['b'] = b
for col_name in self.csv_column:
self.save_fp.write(str(avg_eval_score[col_name]))
if col_name in self.csv_column[:-1]:
self.save_fp.write(',')
elif col_name is self.csv_column[-1]:
self.save_fp.write('\n')
print 'excluded files in evaluation : ', self.excluded_files
k1 = 1.2
b = 0.75
if len(relscore) > 4:
k1 = relscore[3]
b = relscore[4]
ft1 = RelscoreFunctionalTest(
text_dict, data_set_name, relscore[0], int(relscore[1]), int(relscore[2]), float(k1), float(b)
)
# check equality of two methods of getting keyphrase_candidate
for relscore_type in settings["relscore"]:
for winsize in settings["winsize"]:
print "relscore_type : %s, winsize : %s" % (relscore_type, winsize)
old_eval_score = dict()
refactor_eval_score = dict()
old_gs = old_read_gs("/Users/KimKR/Desktop/NEXT_LAB/keyword/gold_standard/%s.ref" % (data_set_name))
refactor_gs = read_gs("/Users/KimKR/Desktop/NEXT_LAB/keyword/gold_standard/%s.ref" % (data_set_name))
for file_name in text_dict.keys():
sentences = text_dict[file_name]
old_relscore = read_relscore(
("/Users/KimKR/Desktop/NEXT_LAB/keyword/rel_score/%s/%s" % (settings["dataset"], relscore_type)),
file_name,
)
refact_relscore = read_relscore("/Users/KimKR/Desktop/NEXT_LAB/keyword/relScore", file_name)
# check equality of keyphrase_candidates
ft = GraphFunctionalTest(file_name, sentences, old_relscore, refact_relscore, int(winsize))
# old_gs[file_name] = old_check_gold_standard(sentences, old_gs[file_name])
# refactor_gs[file_name] = check_gold_standard(sentences, refactor_gs[file_name])
old_eval_score[file_name] = evaluate(ft.kokako_graph.score_candidates(), old_gs[file_name])
refactor_eval_score[file_name] = evaluate(
ft.refactored_graph.score_candidates(1, ft.lamb), refactor_gs[file_name]
)