PARADE/evaluation.py at master · canjiali/PARADE · GitHub

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

import os
import re
import argparse
import subprocess

from scipy import stats
from collections import OrderedDict


parent_path = '/data/tool'
trec_eval_script_path = os.path.join(parent_path, 'trec_eval-9.0.7/trec_eval')
sample_eval_script_path = os.path.join(parent_path, "sample_eval.pl")
gd_eval_script_path = os.path.join(parent_path, "gdeval.pl")


def run(command, get_ouput=False):
  try:
    if get_ouput:
      process = subprocess.Popen(command, stdout=subprocess.PIPE)
      output, err = process.communicate()
      exit_code = process.wait()
      return output
    else:
      subprocess.call(command)
  except subprocess.CalledProcessError as e:
    print(e)


def evaluate_trec(qrels, res, metrics):

  # all_trecs
  command = [trec_eval_script_path, '-m', 'all_trec', '-M', '1000', qrels, res]
  output = run(command, get_ouput=True)
  metrics_val = []
  for metric in metrics:
    metrics_val.append(re.findall(r'{0}\s+all.+\d+'.format(metric), output)[0].split('\t')[2].strip())

  return OrderedDict(zip(metrics, metrics_val))


def evaluate_sample_trec(qrels, res, metrics):
  command = [sample_eval_script_path, qrels, res]
  output = run(command, get_ouput=True)
  metrics_val = []
  for metric in metrics:
    metrics_val.append(re.findall(r'{0}\s+all.+\d+'.format(metric), output)[0].split('\t')[4].strip())

  return OrderedDict(zip(metrics, metrics_val))


def evaluate_metrics(qrels, res, sample_qrels=None, metrics=None):
  normal_metrics = [met for met in metrics if not met.startswith('i')]
  infer_metrics = [met for met in metrics if met.startswith('i')]

  metrics_val_dict = OrderedDict()
  if len(normal_metrics) > 0:
    metrics_val_dict.update(evaluate_trec(qrels, res, metrics=normal_metrics))
  if len(infer_metrics) > 0:
    metrics_val_dict.update(evaluate_sample_trec(sample_qrels, res, metrics=infer_metrics))

  return metrics_val_dict


################################## perquery information ####################################
def evaluate_trec_perquery(qrels, res, metrics):

  # all_trecs
  command = [trec_eval_script_path, '-m', 'all_trec', '-q', '-M', '1000', qrels, res]
  output = run(command, get_ouput=True)
  metrics_val = []
  for metric in metrics:
    curr_res = re.findall(r'{0}\s+\t\d+.+\d+'.format(metric), output)
    curr_res = map(lambda x: float(x.split('\t')[-1]), curr_res)
    metrics_val.append(curr_res)

  return OrderedDict(zip(metrics, metrics_val))


def evaluate_sample_trec_perquery(qrels, res, metrics):
  command = [sample_eval_script_path, '-q', qrels, res]
  output = run(command, get_ouput=True)
  metrics_val = []
  for metric in metrics:
    curr_res = re.findall(r'{0}\s+\t\d+.+\d+'.format(metric), output)
    curr_res = map(lambda x: float(x.split('\t')[-1]), curr_res)
    metrics_val.append(curr_res)

  return OrderedDict(zip(metrics, metrics_val))


def evaluate_metrics_perquery(qrels, res, sample_qrels=None, metrics=None):
  normal_metrics = [met for met in metrics if not met.startswith('i')]
  infer_metrics = [met for met in metrics if met.startswith('i')]
  metrics_val_dict = OrderedDict()
  if len(normal_metrics) > 0:
    metrics_val_dict.update(evaluate_trec_perquery(qrels, res, metrics=normal_metrics))
  if len(infer_metrics) > 0:
    metrics_val_dict.update(evaluate_sample_trec_perquery(sample_qrels, res, metrics=infer_metrics))

  return metrics_val_dict


def tt_test(qrels, res1, res2, sample_qrels=None, metrics=None):
  met_dict1 = evaluate_metrics_perquery(qrels, res1, sample_qrels, metrics)
  met_dict2 = evaluate_metrics_perquery(qrels, res2, sample_qrels, metrics)

  avg_met_dict1 = evaluate_metrics(qrels, res1, sample_qrels, metrics)
  avg_met_dict2 = evaluate_metrics(qrels, res2, sample_qrels, metrics)
  print(avg_met_dict1)
  print(avg_met_dict2)

  test_dict = OrderedDict()
  for met in met_dict1.keys():
    p_value = stats.ttest_rel(met_dict1.get(met), met_dict2.get(met))[1]
    test_dict.update({met: p_value})
  print(test_dict)

  return test_dict

if __name__ == '__main__':

  parser = argparse.ArgumentParser()
  parser.add_argument("--qrels", help="TREC qrels file path")
  parser.add_argument("--baselines", help="Baseline file paths, seperated by ','")
  parser.add_argument("--runs", help="competitive run paths, seperated by ','")
  args = parser.parse_args()

  baselines = args.baselines.split(",")
  runs = args.runs.split(",")
  for trec_run in runs:
    for baseline in baselines:
      print(baseline)
      print(trec_run)
      tt_test(args.qrels, baseline, trec_run, None, ['P_20', 'ndcg_cut_20'])