-
Notifications
You must be signed in to change notification settings - Fork 4
/
caselaw_eval_bm25.py
227 lines (181 loc) · 10.2 KB
/
caselaw_eval_bm25.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
import os
import pytrec_eval
import seaborn as sns
import pickle
sns.set(color_codes=True, font_scale=1.2)
from preprocessing.caselaw_stat_corpus import preprocess_label_file
from eval.eval_bm25 import analyze_correlations_bet_para, aggregate_run_ranks_overlap, aggregate_run_mean_score, \
aggregate_run_overlap, aggregate_run_interleave
from eval.eval_bm25_coliee2021 import aggregate_run_rrf
def eval_ranking_bm25(label_file, bm25_folder, output_dir, output_file: str, measurements, aggregation='interleave', scores='ranks'):
if aggregation == 'overlap_scores':
scores = 'scores'
qrels = preprocess_label_file(label_file)
qrels_updated = {}
for key, value in qrels.items():
qrels_updated.update({key: {}})
for val in value:
qrels_updated.get(str(key)).update({str(val): 1})
if 'separate' in bm25_folder:
print('i do separate')
run = read_run_separate_aggregate(bm25_folder, aggregation, scores)
else:
print('i do whole doc')
run = read_run_whole_doc(bm25_folder, scores)
return run, qrels_updated
def read_run_whole_doc(bm25_folder: str, scores='ranks'):
# geh in den bm25 folder, lies in dokument und query: dann dict {query: {top 1000}}
run = {}
for root, dirs, files in os.walk(bm25_folder):
for file in files:
with open(os.path.join(bm25_folder, file), 'r') as f:
lines = f.readlines()
lines_dict = {}
for i in range(len(lines)):
if scores == 'scores':
lines_dict.update({lines[i].split(' ')[0].strip().strip('_0'): float(lines[i].split(' ')[-1].strip())})
else:
lines_dict.update({lines[i].split(' ')[0].strip().strip('_0'): float(len(lines) - i)})
run.update({file.split('_')[2]: lines_dict})
return run
def read_run_separate(bm25_folder: str, scores='ranks'):
run = {}
for root, dirs, files in os.walk(bm25_folder):
for file in files:
with open(os.path.join(bm25_folder, file), 'r') as f:
lines = f.readlines()
lines_dict = {}
for i in range(len(lines)):
if scores == 'scores':
lines_dict.update({lines[i].split(' ')[0].strip().split('_')[0]: float(lines[i].split(' ')[-1].strip())})
else:
lines_dict.update({lines[i].split(' ')[0].strip().split('_')[0]: len(lines) - i})
if run.get(file.split('_')[2]):
run.get(file.split('_')[2]).update({file.split('_')[3]: lines_dict})
else:
run.update({file.split('_')[2]: {}})
run.get(file.split('_')[2]).update({file.split('_')[3]: lines_dict})
return run
def read_run_separate_aggregate(bm25_folder: str, aggregation='interleave', scores='ranks'):
# geh in den bm25 folder, lies in dokument und query: dann dict {query: {top 1000}}
if aggregation == 'overlap_scores' or aggregation == 'mean_scores':
scores = 'scores'
run = read_run_separate(bm25_folder, scores)
# now i need an aggregation function here, different choices
if aggregation == 'overlap_docs':
# now aggregate according to the overlap of the docs in the paragraphs!
run_aggregated = aggregate_run_overlap(run)
elif aggregation == 'interleave':
run_aggregated = aggregate_run_interleave(run)
elif aggregation == 'overlap_ranks':
# now aggregate according to the overlap of the docs in the paragraphs!
run_aggregated = aggregate_run_ranks_overlap(run)
elif aggregation == 'overlap_scores':
run_aggregated = aggregate_run_ranks_overlap(run)
elif aggregation == 'mean_scores':
run_aggregated = aggregate_run_mean_score(run)
elif aggregation == 'rrf':
run_aggregated = aggregate_run_rrf(run)
if run_aggregated:
return run_aggregated
else:
return run
def ranking_eval(qrels, run, output_dir, measurements, output_file='eval_bm25_aggregate_overlap.txt'):
evaluator = pytrec_eval.RelevanceEvaluator(qrels, measurements)
# measurements)
# {'recall_1', 'recall_2', 'recall_3', 'recall_4', 'recall_5', 'recall_6', 'recall_7', 'recall_8',
# 'recall_9', 'recall_10','recall_11', 'recall_12', 'recall_13', 'recall_14', 'recall_15', 'recall_16', 'recall_17', 'recall_18',
# 'recall_19', 'recall_20','P_1', 'P_2', 'P_3', 'P_4', 'P_5', 'P_6', 'P_7', 'P_8', 'P_9', 'P_10',
# 'P_11', 'P_12', 'P_13', 'P_14', 'P_15', 'P_16', 'P_17', 'P_18', 'P_19', 'P_20'}) # {'recall_100', 'recall_200', 'recall_300', 'recall_500', 'recall_1000'})
results = evaluator.evaluate(run)
def print_line(measure, scope, value):
print('{:25s}{:8s}{:.4f}'.format(measure, scope, value))
def write_line(measure, scope, value):
return '{:25s}{:8s}{:.4f}'.format(measure, scope, value)
for query_id, query_measures in sorted(results.items()):
for measure, value in sorted(query_measures.items()):
print_line(measure, query_id, value)
# for measure in sorted(query_measures.keys()):
# print_line(
# measure,
# 'all',
# pytrec_eval.compute_aggregated_measure(
# measure,
# [query_measures[measure]
# for query_measures in results.values()]))
with open(os.path.join(output_dir, output_file), 'w') as output:
for measure in sorted(query_measures.keys()):
output.write(write_line(
measure,
'all',
pytrec_eval.compute_aggregated_measure(
measure,
[query_measures[measure]
for query_measures in results.values()])) + '\n')
def eval_mode(mode, measurements, label_file):
bm25_folder = '/mnt/c/Users/salthamm/Documents/phd/data/caselaw/bm25/search/{}'.format(mode[1])
output_dir = '/mnt/c/Users/salthamm/Documents/phd/data/caselaw/bm25/eval/{}/'.format(mode[1])
if bm25_folder:
if mode[0] == 'train':
run, qrels = eval_ranking_bm25(label_file, bm25_folder, output_dir, measurements, 'eval_bm25_aggregate_{}.txt'.format(mode[2]),
aggregation=mode[2])
return run, qrels
def ranking_eval2(qrels, run, output_dir, output_file, measurements):
evaluator = pytrec_eval.RelevanceEvaluator(qrels, measurements)
# measurements)
# {'recall_1', 'recall_2', 'recall_3', 'recall_4', 'recall_5', 'recall_6', 'recall_7', 'recall_8',
# 'recall_9', 'recall_10','recall_11', 'recall_12', 'recall_13', 'recall_14', 'recall_15', 'recall_16', 'recall_17', 'recall_18',
# 'recall_19', 'recall_20','P_1', 'P_2', 'P_3', 'P_4', 'P_5', 'P_6', 'P_7', 'P_8', 'P_9', 'P_10',
# 'P_11', 'P_12', 'P_13', 'P_14', 'P_15', 'P_16', 'P_17', 'P_18', 'P_19', 'P_20'}) # {'recall_100', 'recall_200', 'recall_300', 'recall_500', 'recall_1000'})
results = evaluator.evaluate(run)
#bm25_folder = '/mnt/c/Users/salthamm/Documents/phd/data/clef-ip/2011_prior_candidate_search/bm25/search/{}/{}'.format(
# mode[1], mode[0])
#output_dir = '/mnt/c/Users/salthamm/Documents/phd/data/clef-ip/2011_prior_candidate_search/bm25/eval/{}/{}'.format(
# mode[1], mode[0])
#output_file = 'test.txt'
def print_line(measure, scope, value):
print('{:25s}{:8s}{:.4f}'.format(measure, scope, value))
def write_line(measure, scope, value):
return '{:25s}{:8s}{:.4f}'.format(measure, scope, value)
for query_id, query_measures in sorted(results.items()):
for measure, value in sorted(query_measures.items()):
print_line(measure, query_id, value)
# for measure in sorted(query_measures.keys()):
# print_line(
# measure,
# 'all',
# pytrec_eval.compute_aggregated_measure(
# measure,
# [query_measures[measure]
# for query_measures in results.values()]))
with open(os.path.join(output_dir, output_file), 'w') as output:
for measure in sorted(query_measures.keys()):
output.write(write_line(
measure,
'all',
pytrec_eval.compute_aggregated_measure(
measure,
[query_measures[measure]
for query_measures in results.values()])) + '\n')
def eval_mode2(mode, measurements, label_file):
run, qrels = eval_mode(mode, measurements, label_file)
output_dir = '/mnt/c/Users/salthamm/Documents/phd/data/caselaw/bm25/eval/{}'.format(
mode[1])
output_file = 'eval_bm25_{}_{}_aggregate_{}.txt'.format(mode[0], mode[1], mode[2])
ranking_eval2(qrels, run, output_dir, output_file, measurements)
return run, qrels
if __name__ == "__main__":
measurements = {'recall_100', 'recall_200', 'recall_300', 'recall_500', 'recall_1000', 'ndcg_cut_10', 'recip_rank'}
label_file = '/mnt/c/Users/salthamm/Documents/coding/ussc-caselaw-collection/airs2017-collection/qrel.txt'
#run2, qrels2 = eval_mode2(['train', 'whole_doc', 'overlap_docs'], measurements, label_file)
#output_dir = '/mnt/c/Users/salthamm/Documents/phd/data/caselaw/bm25/eval'
#with open(os.path.join(output_dir, 'run_bm25_aggregate2_doc_overlap_ranks.pickle'), 'wb') as f:
# pickle.dump(run2, f)
run, qrel = eval_mode2(['train', 'separate_para', 'rrf'], measurements, label_file)
output_dir = '/mnt/c/Users/salthamm/Documents/phd/data/caselaw/bm25/eval'
with open(os.path.join(output_dir, 'run_bm25_aggregate2_rrf_overlap_ranks.pickle'), 'wb') as f:
pickle.dump(run, f)
#eval_mode2(['train', 'separate_para', 'overlap_docs'], measurements, label_file)
#eval_mode2(['train', 'separate_para', 'overlap_scores'], measurements, label_file)
#eval_mode2(['train', 'separate_para', 'mean_scores'], measurements, label_file)
#eval_mode2(['train', 'separate_para', 'interleave'], measurements, label_file)