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
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420 | class MECOProcessor(DatasetProcessor):
"""Processor MECO dataset"""
def __init__(self, *args, **kwargs):
super().__init__(*args, **kwargs)
self._nlp = None
self._surp_extractor = None
def dataset_specific_processing(
self, data_dict: dict[str, pd.DataFrame]
) -> dict[str, pd.DataFrame]:
"""MECO-specific processing steps"""
# add unique trial IDs and merge labels
labels = self._load_labels().rename(columns={'uniform_id': Fields.SUBJECT_ID})
for data_type in [DataType.IA, DataType.FIXATIONS]:
df = data_dict[data_type]
df[Fields.UNIQUE_TRIAL_ID] = (
df[Fields.SUBJECT_ID].astype(str)
+ '_'
+ df[Fields.UNIQUE_PARAGRAPH_ID].astype(str)
)
df = df.merge(labels, on='participant_id', validate='many_to_one')
df = df[~df['lextale'].isna()]
data_dict[data_type] = df
# add IA features to fixations
data_dict['fixations'], data_dict['ia'] = (
self.add_ia_report_features_to_fixation_data(
data_dict['ia'],
data_dict['fixations'],
)
)
# add missing features
for data_type in [DataType.IA, DataType.FIXATIONS]:
data_dict[data_type] = add_missing_features(
et_data=data_dict[data_type],
trial_groupby_columns=self.data_args.groupby_columns,
mode=data_type,
)
data_dict[data_type] = data_dict[data_type].assign(
normalized_ID=(
data_dict[data_type]['IA_ID'] - data_dict[data_type]['IA_ID'].min()
)
/ (
data_dict[data_type]['IA_ID'].max()
- data_dict[data_type]['IA_ID'].min()
),
)
# compute trial-level features
trial_level_features = compute_trial_level_features(
raw_fixation_data=data_dict[DataType.FIXATIONS],
raw_ia_data=data_dict[DataType.IA],
trial_groupby_columns=self.data_args.groupby_columns,
processed_data_path=self.data_args.processed_data_path,
)
data_dict[DataType.TRIAL_LEVEL] = trial_level_features
# replace missing values
data_dict = replace_missing_values(data_dict)
return data_dict
@property
def nlp(self):
if self._nlp is None:
self._nlp = spacy.load('en_core_web_sm')
return self._nlp
@property
def surp_extractor(self):
if self._surp_extractor is None:
self._surp_extractor = get_surp_extractor(
extractor_type=SurpExtractorType.CAT_CTX_LEFT, model_name='gpt2'
)
return self._surp_extractor
def _prepare_ia_dataframe(self, ia_df: pd.DataFrame) -> pd.DataFrame:
ia_df = ia_df.rename(
columns={
Fields.IA_DATA_IA_ID_COL_NAME: Fields.FIXATION_REPORT_IA_ID_COL_NAME
}
)
# make sure order preserved
ia_df = ia_df.sort_values(
['unique_trial_id', 'CURRENT_FIX_INTEREST_AREA_INDEX']
)
# group operations in batch
grouped = ia_df.groupby('unique_trial_id')
ia_df['TRIAL_IA_COUNT'] = grouped['unique_trial_id'].transform('count')
ia_df['IA_DWELL_TIME_%'] = grouped['IA_DWELL_TIME'].transform(
lambda x: x / np.sum(x)
)
ia_df['PARAGRAPH_RT'] = ia_df.groupby(Fields.UNIQUE_PARAGRAPH_ID)[
'IA_DWELL_TIME'
].transform('sum')
ia_df['IA_FIXATION_%'] = ia_df.groupby('unique_trial_id')[
'IA_FIXATION_COUNT'
].transform(lambda x: x / x.sum())
# feature calculations
ia_df['IA_FIRST_FIX_PROGRESSIVE'] = (ia_df['firstfix.sac.in'] > 0).astype(int)
ia_df['IA_RUN_COUNT'] = ia_df['nrun']
ia_df['IA_SELECTIVE_REGRESSION_PATH_DURATION'] = ia_df['firstrun.gopast.sel']
ia_df['IA_SKIP'] = ia_df['total_skip']
ia_df['IA_FIRST_FIX_PROGRESSIVE'] = (ia_df['firstfix.sac.in'] > 0).astype(int)
ia_df['word_length'] = ia_df['IA_LABEL'].str.len()
# not really the feature but better than 0, it is a binary indicator only
ia_df['IA_REGRESSION_IN_COUNT'] = ia_df['IA_REGRESSION_IN']
ia_df['IA_REGRESSION_OUT_FULL_COUNT'] = ia_df['IA_REGRESSION_OUT']
ia_df['IA_REGRESSION_OUT_COUNT'] = ia_df['IA_REGRESSION_OUT']
ia_df['IA_ID'] = ia_df[Fields.FIXATION_REPORT_IA_ID_COL_NAME]
# missing columns
zero_cols = [
'NEXT_FIX_INTEREST_AREA_INDEX',
'CURRENT_FIX_NEAREST_INTEREST_AREA_DISTANCE',
'start_of_line',
'end_of_line',
'IA_LAST_FIXATION_DURATION',
'IA_LAST_RUN_DWELL_TIME',
'IA_REGRESSION_PATH_DURATION',
'IA_LAST_RUN_FIXATION_COUNT',
'IA_FIRST_FIXATION_VISITED_IA_COUNT',
'IA_LEFT',
'IA_RIGHT',
'IA_TOP',
'IA_BOTTOM',
'NEXT_SAC_DURATION',
'NEXT_SAC_AVG_VELOCITY',
'NEXT_SAC_AMPLITUDE',
'NEXT_SAC_END_X',
'NEXT_SAC_START_X',
'NEXT_SAC_START_Y',
'NEXT_SAC_END_Y',
]
ia_df[zero_cols] = 0
return ia_df
def _process_metrics_batch(self, groups: list[pd.DataFrame]) -> pd.DataFrame:
metrics_list = []
for group in tqdm(groups, desc='Sequential metric extraction'):
try:
sentence = group['paragraph'].iloc[0]
metrics = get_metrics(
target_text=sentence,
surp_extractor=self.surp_extractor,
parsing_model=self.nlp,
parsing_mode='re-tokenize',
add_parsing_features=True,
language='en',
)
metrics['unique_paragraph_id'] = group['unique_paragraph_id'].iloc[0]
metrics['participant_id'] = group['participant_id'].iloc[0]
metrics[Fields.FIXATION_REPORT_IA_ID_COL_NAME] = (
metrics['Token_idx'] + 1
)
metrics_list.append(metrics)
except Exception as e:
logger.warning(
f'Error processing group {group["unique_paragraph_id"].iloc[0]}: {e}'
)
continue
return (
pd.concat(metrics_list, ignore_index=True)
if metrics_list
else pd.DataFrame()
)
def _merge_metrics_to_ia(
self, ia_df: pd.DataFrame, metrics_df: pd.DataFrame
) -> pd.DataFrame:
merge_keys = ['unique_trial_id', Fields.FIXATION_REPORT_IA_ID_COL_NAME]
metrics_df[Fields.UNIQUE_TRIAL_ID] = (
metrics_df[Fields.SUBJECT_ID].astype(str)
+ '_'
+ metrics_df[Fields.UNIQUE_PARAGRAPH_ID].astype(str)
)
drop_keys = (set(metrics_df.columns) & set(ia_df.columns)) - set(merge_keys)
cols_to_drop = list(drop_keys) + ['Morph']
cols_to_drop = [c for c in cols_to_drop if c in metrics_df.columns]
metrics_clean = metrics_df.drop(columns=cols_to_drop).drop_duplicates()
ia_df = ia_df.merge(
metrics_clean,
on=merge_keys,
how='left',
validate='many_to_one',
)
rename_map = {
'POS': 'universal_pos',
'Length': 'word_length_no_punctuation',
'Wordfreq_Frequency': 'wordfreq_frequency',
'subtlex_Frequency': 'subtlex_frequency',
'Reduced_POS': 'ptb_pos',
'Head_word_idx': 'head_word_index',
'Dependency_Relation': 'dependency_relation',
'Entity': 'entity_type',
'gpt2_Surprisal': 'gpt2_surprisal',
'gpt2': 'gpt2_surprisal',
'Head_Direction': 'head_direction',
'Is_Content_Word': 'is_content_word',
'n_Lefts': 'left_dependents_count',
'n_Rights': 'right_dependents_count',
'Distance2Head': 'distance_to_head',
}
ia_df = ia_df.rename(columns=rename_map)
return ia_df
def add_ia_report_features_to_fixation_data(
self,
ia_df: pd.DataFrame,
fix_df: pd.DataFrame,
) -> tuple[pd.DataFrame, pd.DataFrame]:
# Deduplicate groupby columns
self.data_args.groupby_columns = list(
dict.fromkeys(self.data_args.groupby_columns)
)
ia_df = self._prepare_ia_dataframe(ia_df)
# fix misalignments in meco
# wrong paragraph ids for participant sp_36 from 5 to 12 (off by one)
key_value_map = {
'sp_36': {11: 12, 10: 11, 9: 10, 8: 9, 7: 8, 6: 7, 5: 6},
'gr_45': {11: 12, 10: 11, 9: 10, 8: 9, 7: 8},
'it_25': {10: 11, 9: 10, 7: 8, 6: 7},
'se_38': {11: 12, 10: 11, 9: 10, 8: 9, 7: 8, 6: 7, 5: 6, 4: 5},
}
for participant_id, mapping in key_value_map.items():
mask = ia_df.participant_id == participant_id
fix_mask = fix_df.participant_id == participant_id
for old_id, new_id in sorted(mapping.items(), reverse=True):
ia_df.loc[
mask & (ia_df.unique_paragraph_id == old_id), 'unique_paragraph_id'
] = new_id
fix_df.loc[
fix_mask & (fix_df.unique_paragraph_id == old_id),
'unique_paragraph_id',
] = new_id
ia_df['unique_trial_id'] = (
ia_df[Fields.SUBJECT_ID].astype(str)
+ '_'
+ ia_df[Fields.UNIQUE_PARAGRAPH_ID].astype(str)
)
fix_df['unique_trial_id'] = (
fix_df[Fields.SUBJECT_ID].astype(str)
+ '_'
+ fix_df[Fields.UNIQUE_PARAGRAPH_ID].astype(str)
)
# load stimuli because ia_df seems to contain different number of aois
# in this way we can be sure to that all paragraphs are what MECO authors provide
stimuli_df = pd.read_csv(
'data/MECOL2/stimuli/stimuli.csv', engine='python', encoding='latin1'
)
stimuli_df = stimuli_df.rename(
columns={'trialid': 'unique_paragraph_id', 'text': 'paragraph'},
)
stimuli_df = stimuli_df.drop(columns=['question1', 'question2'])
ia_df = ia_df.merge(
stimuli_df,
on='unique_paragraph_id',
validate='many_to_one',
)
fix_df = fix_df.merge(
stimuli_df,
on='unique_paragraph_id',
validate='many_to_one',
)
groups = [group for _, group in ia_df.groupby('unique_paragraph_id')]
metrics_df = self._process_metrics_batch(groups)
ia_df = self._merge_metrics_to_ia(ia_df, metrics_df)
fix_df['NEXT_FIX_INTEREST_AREA_INDEX'] = fix_df[
'CURRENT_FIX_INTEREST_AREA_INDEX'
].shift(-1)
fix_df['CURRENT_FIX_INTEREST_AREA_INDEX'] = fix_df[
'CURRENT_FIX_INTEREST_AREA_INDEX'
].fillna(-1)
if 'normalized_part_ID' in fix_df.columns:
if fix_df['normalized_part_ID'].isna().any():
logger.info('normalized_part_ID contains NaNs; dropping it.')
fix_df = fix_df.drop(columns='normalized_part_ID')
merge_keys = self.data_args.groupby_columns + [
Fields.FIXATION_REPORT_IA_ID_COL_NAME
]
dup_cols = (set(fix_df.columns) & set(ia_df.columns)) - set(merge_keys)
_ia_df = ia_df.drop(columns=list(dup_cols))
enriched_fix_df = fix_df.merge(
_ia_df.drop_duplicates(subset=merge_keys, keep='first'),
on=merge_keys,
how='left',
validate='many_to_one',
)
num_of_words = ia_df.groupby(self.data_args.groupby_columns).size()
num_of_words.name = 'num_of_words_in_trial'
enriched_fix_df = enriched_fix_df.merge(
num_of_words,
on=self.data_args.groupby_columns,
how='left',
validate='many_to_one',
)
# convert types
float_cols = ['TOWRE_word', 'TOWRE_nonword']
for col in float_cols:
if col in enriched_fix_df.columns:
enriched_fix_df[col] = enriched_fix_df[col].astype(float)
if col in ia_df.columns:
ia_df[col] = ia_df[col].astype(float)
enriched_fix_df['TRIAL_IA_COUNT'] = enriched_fix_df['TRIAL_IA_COUNT'].fillna(0)
return enriched_fix_df, ia_df
def get_column_map(self, data_type: DataType) -> dict:
column_maps = {
DataType.IA: {
'uniform_id': Fields.SUBJECT_ID,
'trialid': Fields.UNIQUE_PARAGRAPH_ID,
'skip': 'total_skip',
'wordnum': 'IA_ID',
'word': 'IA_LABEL',
'nfix': 'IA_FIXATION_COUNT',
'reg.in': 'IA_REGRESSION_IN',
'reg.out': 'IA_REGRESSION_OUT',
'dur': 'IA_DWELL_TIME',
'firstrun.nfix': 'IA_FIRST_RUN_FIXATION_COUNT',
'firstrun.dur': 'IA_FIRST_RUN_DWELL_TIME',
'firstfix.launch': 'IA_FIRST_RUN_LAUNCH_SITE',
'firstfix.land': 'IA_FIRST_RUN_LANDING_POSITION',
'firstfix.dur': 'IA_FIRST_FIXATION_DURATION',
},
DataType.FIXATIONS: {
'xn': 'CURRENT_FIX_X',
'yn': 'CURRENT_FIX_Y',
'dur': 'CURRENT_FIX_DURATION',
'uniform_id': Fields.SUBJECT_ID,
'trialid': Fields.UNIQUE_PARAGRAPH_ID,
'fixid': 'CURRENT_FIX_INDEX',
'start': 'CURENT_FIX_START',
'stop': 'CURRENT_FIX_END',
'ps': 'CURRENT_FIX_PUPIL_SIZE',
'blink': 'CURRENT_FIX_BLINK_AROUND',
'word': 'CURRENT_FIX_INTEREST_AREA_LABEL',
'ianum': 'CURRENT_FIX_INTEREST_AREA_INDEX',
'ia': 'CURRENT_FIX_LABEL',
'ia.fix': 'CURRENT_FIX_INTEREST_AREA_FIX_COUNT',
'ia.runid': 'CURRENT_FIX_INTEREST_AREA_ID',
},
}
return column_maps.get(data_type, {})
def get_columns_to_keep(self) -> list:
"""Get list of columns to keep after filtering"""
return []
@staticmethod
@lru_cache(maxsize=2)
def _load_labels():
labels_w1 = pyreadr.read_r(
'data/MECOL2W1/demographics/joint.ind.diff.l2.rda',
)['joint_id']
labels_w2 = pyreadr.read_r(
'data/MECOL2W2/demographics/joint.ind.diff.l2.w2.rda',
)['joint_id_w2']
return pd.concat([labels_w1, labels_w2], axis=0).reset_index(drop=True)
|