illorca commited on
Commit
d8424e9
1 Parent(s): 79a6fc4

Include weighted mode. ORIGINAL FAIREVAL SCRIPT IS MODIFIED

Browse files
Files changed (2) hide show
  1. FairEval.py +23 -7
  2. FairEvalUtils.py +2 -1
FairEval.py CHANGED
@@ -119,6 +119,7 @@ class FairEvaluation(evaluate.Metric):
119
  suffix: bool = False,
120
  scheme: Optional[str] = None,
121
  mode: Optional[str] = 'fair',
 
122
  error_format: Optional[str] = 'count',
123
  zero_division: Union[str, int] = "warn",
124
  ):
@@ -147,25 +148,38 @@ class FairEvaluation(evaluate.Metric):
147
  pred_spans = seq_to_fair(pred_spans)
148
 
149
  # (3) COUNT ERRORS AND CALCULATE SCORES
150
- total_errors = compare_spans([], []) # initialize empty error count dictionary
151
-
152
  for i in range(len(true_spans)):
153
  sentence_errors = compare_spans(true_spans[i], pred_spans[i])
154
  total_errors = add_dict(total_errors, sentence_errors)
155
 
156
- results = calculate_results(total_errors)
 
 
 
 
 
 
 
 
 
 
 
157
  del results['conf']
158
 
159
- # (4) SELECT OUTPUT MODE AND REFORMAT AS SEQEVAL HUGGINGFACE OUTPUT
 
160
  output = {}
161
  total_trad_errors = results['overall']['traditional']['FP'] + results['overall']['traditional']['FN']
162
  total_fair_errors = results['overall']['fair']['FP'] + results['overall']['fair']['FN'] + \
163
  results['overall']['fair']['LE'] + results['overall']['fair']['BE'] + \
164
  results['overall']['fair']['LBE']
165
 
166
- assert mode in ['traditional', 'fair'], 'mode must be \'traditional\' or \'fair\''
 
167
  assert error_format in ['count', 'proportion'], 'error_format must be \'count\' or \'proportion\''
168
 
 
169
  if mode == 'traditional':
170
  for k, v in results['per_label'][mode].items():
171
  if error_format == 'count':
@@ -174,7 +188,7 @@ class FairEvaluation(evaluate.Metric):
174
  elif error_format == 'proportion':
175
  output[k] = {'precision': v['Prec'], 'recall': v['Rec'], 'f1': v['F1'], 'TP': v['TP'],
176
  'FP': v['FP'] / total_trad_errors, 'FN': v['FN'] / total_trad_errors}
177
- elif mode == 'fair':
178
  for k, v in results['per_label'][mode].items():
179
  if error_format == 'count':
180
  output[k] = {'precision': v['Prec'], 'recall': v['Rec'], 'f1': v['F1'], 'TP': v['TP'],
@@ -185,10 +199,12 @@ class FairEvaluation(evaluate.Metric):
185
  'LE': v['LE'] / total_fair_errors, 'BE': v['BE'] / total_fair_errors,
186
  'LBE': v['LBE'] / total_fair_errors}
187
 
 
188
  output['overall_precision'] = results['overall'][mode]['Prec']
189
  output['overall_recall'] = results['overall'][mode]['Rec']
190
  output['overall_f1'] = results['overall'][mode]['F1']
191
 
 
192
  if mode == 'traditional':
193
  output['TP'] = results['overall'][mode]['TP']
194
  output['FP'] = results['overall'][mode]['FP']
@@ -196,7 +212,7 @@ class FairEvaluation(evaluate.Metric):
196
  if error_format == 'proportion':
197
  output['FP'] = output['FP'] / total_trad_errors
198
  output['FN'] = output['FN'] / total_trad_errors
199
- elif mode == 'fair':
200
  output['TP'] = results['overall'][mode]['TP']
201
  output['FP'] = results['overall'][mode]['FP']
202
  output['FN'] = results['overall'][mode]['FN']
 
119
  suffix: bool = False,
120
  scheme: Optional[str] = None,
121
  mode: Optional[str] = 'fair',
122
+ weights: dict = None,
123
  error_format: Optional[str] = 'count',
124
  zero_division: Union[str, int] = "warn",
125
  ):
 
148
  pred_spans = seq_to_fair(pred_spans)
149
 
150
  # (3) COUNT ERRORS AND CALCULATE SCORES
151
+ total_errors = compare_spans([], [])
 
152
  for i in range(len(true_spans)):
153
  sentence_errors = compare_spans(true_spans[i], pred_spans[i])
154
  total_errors = add_dict(total_errors, sentence_errors)
155
 
156
+ if weights is None and mode == 'weighted':
157
+ print("The chosen mode is \'weighted\', but no weights are given. Setting weights to:\n")
158
+ weights = {"TP": {"TP": 1},
159
+ "FP": {"FP": 1},
160
+ "FN": {"FN": 1},
161
+ "LE": {"TP": 0, "FP": 0.5, "FN": 0.5},
162
+ "BE": {"TP": 0.5, "FP": 0.25, "FN": 0.25},
163
+ "LBE": {"TP": 0, "FP": 0.5, "FN": 0.5}}
164
+ print(weights)
165
+
166
+ config = {"labels": "all", "eval_method": [mode], "weights": weights,}
167
+ results = calculate_results(total_errors, config)
168
  del results['conf']
169
 
170
+ # (4) SELECT OUTPUT MODE AND REFORMAT AS SEQEVAL-HUGGINGFACE OUTPUT
171
+ # initialize empty dictionary and count errors
172
  output = {}
173
  total_trad_errors = results['overall']['traditional']['FP'] + results['overall']['traditional']['FN']
174
  total_fair_errors = results['overall']['fair']['FP'] + results['overall']['fair']['FN'] + \
175
  results['overall']['fair']['LE'] + results['overall']['fair']['BE'] + \
176
  results['overall']['fair']['LBE']
177
 
178
+ # assert valid options
179
+ assert mode in ['traditional', 'fair', 'weighted'], 'mode must be \'traditional\', \'fair\' or \'weighted\''
180
  assert error_format in ['count', 'proportion'], 'error_format must be \'count\' or \'proportion\''
181
 
182
+ # append entity-level errors and scores
183
  if mode == 'traditional':
184
  for k, v in results['per_label'][mode].items():
185
  if error_format == 'count':
 
188
  elif error_format == 'proportion':
189
  output[k] = {'precision': v['Prec'], 'recall': v['Rec'], 'f1': v['F1'], 'TP': v['TP'],
190
  'FP': v['FP'] / total_trad_errors, 'FN': v['FN'] / total_trad_errors}
191
+ elif mode == 'fair' or mode == 'weighted':
192
  for k, v in results['per_label'][mode].items():
193
  if error_format == 'count':
194
  output[k] = {'precision': v['Prec'], 'recall': v['Rec'], 'f1': v['F1'], 'TP': v['TP'],
 
199
  'LE': v['LE'] / total_fair_errors, 'BE': v['BE'] / total_fair_errors,
200
  'LBE': v['LBE'] / total_fair_errors}
201
 
202
+ # append overall scores
203
  output['overall_precision'] = results['overall'][mode]['Prec']
204
  output['overall_recall'] = results['overall'][mode]['Rec']
205
  output['overall_f1'] = results['overall'][mode]['F1']
206
 
207
+ # append overall error counts
208
  if mode == 'traditional':
209
  output['TP'] = results['overall'][mode]['TP']
210
  output['FP'] = results['overall'][mode]['FP']
 
212
  if error_format == 'proportion':
213
  output['FP'] = output['FP'] / total_trad_errors
214
  output['FN'] = output['FN'] / total_trad_errors
215
+ elif mode == 'fair' or 'weighted':
216
  output['TP'] = results['overall'][mode]['TP']
217
  output['FP'] = results['overall'][mode]['FP']
218
  output['FN'] = results['overall'][mode]['FN']
FairEvalUtils.py CHANGED
@@ -1149,7 +1149,7 @@ def add_dict(base_dict, dict_to_add):
1149
 
1150
  #############################
1151
 
1152
- def calculate_results(eval_dict, **config):
1153
  """
1154
  Calculate overall precision, recall, and F-scores.
1155
 
@@ -1173,6 +1173,7 @@ def calculate_results(eval_dict, **config):
1173
  eval_dict["overall"]["weighted"] = {}
1174
  for err_type in eval_dict["overall"]["fair"]:
1175
  eval_dict["overall"]["weighted"][err_type] = eval_dict["overall"]["fair"][err_type]
 
1176
  for label in eval_dict["per_label"]["fair"]:
1177
  eval_dict["per_label"]["weighted"][label] = {}
1178
  for err_type in eval_dict["per_label"]["fair"][label]:
 
1149
 
1150
  #############################
1151
 
1152
+ def calculate_results(eval_dict, config):
1153
  """
1154
  Calculate overall precision, recall, and F-scores.
1155
 
 
1173
  eval_dict["overall"]["weighted"] = {}
1174
  for err_type in eval_dict["overall"]["fair"]:
1175
  eval_dict["overall"]["weighted"][err_type] = eval_dict["overall"]["fair"][err_type]
1176
+ eval_dict["per_label"]["weighted"] = {}
1177
  for label in eval_dict["per_label"]["fair"]:
1178
  eval_dict["per_label"]["weighted"][label] = {}
1179
  for err_type in eval_dict["per_label"]["fair"][label]: