File size: 40,789 Bytes
c4b2b37 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 963 964 965 966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 |
import argparse
import json
import logging
import os
import pprint
from collections import Counter, defaultdict, namedtuple
from dataclasses import dataclass
from itertools import chain
from typing import Any, Callable, Dict, List, Set, Tuple
import numpy as np
import torch
from BERT_rationale_benchmark.utils import (Annotation, Evidence,
annotations_from_jsonl,
load_documents,
load_flattened_documents,
load_jsonl)
from scipy.stats import entropy
from sklearn.metrics import (accuracy_score, auc, average_precision_score,
classification_report, precision_recall_curve,
roc_auc_score)
logging.basicConfig(
level=logging.DEBUG, format="%(relativeCreated)6d %(threadName)s %(message)s"
)
# start_token is inclusive, end_token is exclusive
@dataclass(eq=True, frozen=True)
class Rationale:
ann_id: str
docid: str
start_token: int
end_token: int
def to_token_level(self) -> List["Rationale"]:
ret = []
for t in range(self.start_token, self.end_token):
ret.append(Rationale(self.ann_id, self.docid, t, t + 1))
return ret
@classmethod
def from_annotation(cls, ann: Annotation) -> List["Rationale"]:
ret = []
for ev_group in ann.evidences:
for ev in ev_group:
ret.append(
Rationale(ann.annotation_id, ev.docid, ev.start_token, ev.end_token)
)
return ret
@classmethod
def from_instance(cls, inst: dict) -> List["Rationale"]:
ret = []
for rat in inst["rationales"]:
for pred in rat.get("hard_rationale_predictions", []):
ret.append(
Rationale(
inst["annotation_id"],
rat["docid"],
pred["start_token"],
pred["end_token"],
)
)
return ret
@dataclass(eq=True, frozen=True)
class PositionScoredDocument:
ann_id: str
docid: str
scores: Tuple[float]
truths: Tuple[bool]
@classmethod
def from_results(
cls,
instances: List[dict],
annotations: List[Annotation],
docs: Dict[str, List[Any]],
use_tokens: bool = True,
) -> List["PositionScoredDocument"]:
"""Creates a paired list of annotation ids/docids/predictions/truth values"""
key_to_annotation = dict()
for ann in annotations:
for ev in chain.from_iterable(ann.evidences):
key = (ann.annotation_id, ev.docid)
if key not in key_to_annotation:
key_to_annotation[key] = [False for _ in docs[ev.docid]]
if use_tokens:
start, end = ev.start_token, ev.end_token
else:
start, end = ev.start_sentence, ev.end_sentence
for t in range(start, end):
key_to_annotation[key][t] = True
ret = []
if use_tokens:
field = "soft_rationale_predictions"
else:
field = "soft_sentence_predictions"
for inst in instances:
for rat in inst["rationales"]:
docid = rat["docid"]
scores = rat[field]
key = (inst["annotation_id"], docid)
assert len(scores) == len(docs[docid])
if key in key_to_annotation:
assert len(scores) == len(key_to_annotation[key])
else:
# In case model makes a prediction on docuemnt(s) for which ground truth evidence is not present
key_to_annotation[key] = [False for _ in docs[docid]]
ret.append(
PositionScoredDocument(
inst["annotation_id"],
docid,
tuple(scores),
tuple(key_to_annotation[key]),
)
)
return ret
def _f1(_p, _r):
if _p == 0 or _r == 0:
return 0
return 2 * _p * _r / (_p + _r)
def _keyed_rationale_from_list(
rats: List[Rationale],
) -> Dict[Tuple[str, str], Rationale]:
ret = defaultdict(set)
for r in rats:
ret[(r.ann_id, r.docid)].add(r)
return ret
def partial_match_score(
truth: List[Rationale], pred: List[Rationale], thresholds: List[float]
) -> List[Dict[str, Any]]:
"""Computes a partial match F1
Computes an instance-level (annotation) micro- and macro-averaged F1 score.
True Positives are computed by using intersection-over-union and
thresholding the resulting intersection-over-union fraction.
Micro-average results are computed by ignoring instance level distinctions
in the TP calculation (and recall, and precision, and finally the F1 of
those numbers). Macro-average results are computed first by measuring
instance (annotation + document) precisions and recalls, averaging those,
and finally computing an F1 of the resulting average.
"""
ann_to_rat = _keyed_rationale_from_list(truth)
pred_to_rat = _keyed_rationale_from_list(pred)
num_classifications = {k: len(v) for k, v in pred_to_rat.items()}
num_truth = {k: len(v) for k, v in ann_to_rat.items()}
ious = defaultdict(dict)
for k in set(ann_to_rat.keys()) | set(pred_to_rat.keys()):
for p in pred_to_rat.get(k, []):
best_iou = 0.0
for t in ann_to_rat.get(k, []):
num = len(
set(range(p.start_token, p.end_token))
& set(range(t.start_token, t.end_token))
)
denom = len(
set(range(p.start_token, p.end_token))
| set(range(t.start_token, t.end_token))
)
iou = 0 if denom == 0 else num / denom
if iou > best_iou:
best_iou = iou
ious[k][p] = best_iou
scores = []
for threshold in thresholds:
threshold_tps = dict()
for k, vs in ious.items():
threshold_tps[k] = sum(int(x >= threshold) for x in vs.values())
micro_r = (
sum(threshold_tps.values()) / sum(num_truth.values())
if sum(num_truth.values()) > 0
else 0
)
micro_p = (
sum(threshold_tps.values()) / sum(num_classifications.values())
if sum(num_classifications.values()) > 0
else 0
)
micro_f1 = _f1(micro_r, micro_p)
macro_rs = list(
threshold_tps.get(k, 0.0) / n if n > 0 else 0 for k, n in num_truth.items()
)
macro_ps = list(
threshold_tps.get(k, 0.0) / n if n > 0 else 0
for k, n in num_classifications.items()
)
macro_r = sum(macro_rs) / len(macro_rs) if len(macro_rs) > 0 else 0
macro_p = sum(macro_ps) / len(macro_ps) if len(macro_ps) > 0 else 0
macro_f1 = _f1(macro_r, macro_p)
scores.append(
{
"threshold": threshold,
"micro": {"p": micro_p, "r": micro_r, "f1": micro_f1},
"macro": {"p": macro_p, "r": macro_r, "f1": macro_f1},
}
)
return scores
def score_hard_rationale_predictions(
truth: List[Rationale], pred: List[Rationale]
) -> Dict[str, Dict[str, float]]:
"""Computes instance (annotation)-level micro/macro averaged F1s"""
scores = dict()
truth = set(truth)
pred = set(pred)
micro_prec = len(truth & pred) / len(pred)
micro_rec = len(truth & pred) / len(truth)
micro_f1 = _f1(micro_prec, micro_rec)
scores["instance_micro"] = {
"p": micro_prec,
"r": micro_rec,
"f1": micro_f1,
}
ann_to_rat = _keyed_rationale_from_list(truth)
pred_to_rat = _keyed_rationale_from_list(pred)
instances_to_scores = dict()
for k in set(ann_to_rat.keys()) | (pred_to_rat.keys()):
if len(pred_to_rat.get(k, set())) > 0:
instance_prec = len(
ann_to_rat.get(k, set()) & pred_to_rat.get(k, set())
) / len(pred_to_rat[k])
else:
instance_prec = 0
if len(ann_to_rat.get(k, set())) > 0:
instance_rec = len(
ann_to_rat.get(k, set()) & pred_to_rat.get(k, set())
) / len(ann_to_rat[k])
else:
instance_rec = 0
instance_f1 = _f1(instance_prec, instance_rec)
instances_to_scores[k] = {
"p": instance_prec,
"r": instance_rec,
"f1": instance_f1,
}
# these are calculated as sklearn would
macro_prec = sum(instance["p"] for instance in instances_to_scores.values()) / len(
instances_to_scores
)
macro_rec = sum(instance["r"] for instance in instances_to_scores.values()) / len(
instances_to_scores
)
macro_f1 = sum(instance["f1"] for instance in instances_to_scores.values()) / len(
instances_to_scores
)
f1_scores = [instance["f1"] for instance in instances_to_scores.values()]
print(macro_f1, np.argsort(f1_scores)[::-1])
scores["instance_macro"] = {
"p": macro_prec,
"r": macro_rec,
"f1": macro_f1,
}
return scores
def _auprc(truth: Dict[Any, List[bool]], preds: Dict[Any, List[float]]) -> float:
if len(preds) == 0:
return 0.0
assert len(truth.keys() and preds.keys()) == len(truth.keys())
aucs = []
for k, true in truth.items():
pred = preds[k]
true = [int(t) for t in true]
precision, recall, _ = precision_recall_curve(true, pred)
aucs.append(auc(recall, precision))
return np.average(aucs)
def _score_aggregator(
truth: Dict[Any, List[bool]],
preds: Dict[Any, List[float]],
score_function: Callable[[List[float], List[float]], float],
discard_single_class_answers: bool,
) -> float:
if len(preds) == 0:
return 0.0
assert len(truth.keys() and preds.keys()) == len(truth.keys())
scores = []
for k, true in truth.items():
pred = preds[k]
if (all(true) or all(not x for x in true)) and discard_single_class_answers:
continue
true = [int(t) for t in true]
scores.append(score_function(true, pred))
return np.average(scores)
def score_soft_tokens(paired_scores: List[PositionScoredDocument]) -> Dict[str, float]:
truth = {(ps.ann_id, ps.docid): ps.truths for ps in paired_scores}
pred = {(ps.ann_id, ps.docid): ps.scores for ps in paired_scores}
auprc_score = _auprc(truth, pred)
ap = _score_aggregator(truth, pred, average_precision_score, True)
roc_auc = _score_aggregator(truth, pred, roc_auc_score, True)
return {
"auprc": auprc_score,
"average_precision": ap,
"roc_auc_score": roc_auc,
}
def _instances_aopc(
instances: List[dict], thresholds: List[float], key: str
) -> Tuple[float, List[float]]:
dataset_scores = []
for inst in instances:
kls = inst["classification"]
beta_0 = inst["classification_scores"][kls]
instance_scores = []
for score in filter(
lambda x: x["threshold"] in thresholds,
sorted(inst["thresholded_scores"], key=lambda x: x["threshold"]),
):
beta_k = score[key][kls]
delta = beta_0 - beta_k
instance_scores.append(delta)
assert len(instance_scores) == len(thresholds)
dataset_scores.append(instance_scores)
dataset_scores = np.array(dataset_scores)
# a careful reading of Samek, et al. "Evaluating the Visualization of What a Deep Neural Network Has Learned"
# and some algebra will show the reader that we can average in any of several ways and get the same result:
# over a flattened array, within an instance and then between instances, or over instances (by position) an
# then across them.
final_score = np.average(dataset_scores)
position_scores = np.average(dataset_scores, axis=0).tolist()
return final_score, position_scores
def compute_aopc_scores(instances: List[dict], aopc_thresholds: List[float]):
if aopc_thresholds is None:
aopc_thresholds = sorted(
set(
chain.from_iterable(
[x["threshold"] for x in y["thresholded_scores"]] for y in instances
)
)
)
aopc_comprehensiveness_score, aopc_comprehensiveness_points = _instances_aopc(
instances, aopc_thresholds, "comprehensiveness_classification_scores"
)
aopc_sufficiency_score, aopc_sufficiency_points = _instances_aopc(
instances, aopc_thresholds, "sufficiency_classification_scores"
)
return (
aopc_thresholds,
aopc_comprehensiveness_score,
aopc_comprehensiveness_points,
aopc_sufficiency_score,
aopc_sufficiency_points,
)
def score_classifications(
instances: List[dict],
annotations: List[Annotation],
docs: Dict[str, List[str]],
aopc_thresholds: List[float],
) -> Dict[str, float]:
def compute_kl(cls_scores_, faith_scores_):
keys = list(cls_scores_.keys())
cls_scores_ = [cls_scores_[k] for k in keys]
faith_scores_ = [faith_scores_[k] for k in keys]
return entropy(faith_scores_, cls_scores_)
labels = list(set(x.classification for x in annotations))
label_to_int = {l: i for i, l in enumerate(labels)}
key_to_instances = {inst["annotation_id"]: inst for inst in instances}
truth = []
predicted = []
for ann in annotations:
truth.append(label_to_int[ann.classification])
inst = key_to_instances[ann.annotation_id]
predicted.append(label_to_int[inst["classification"]])
classification_scores = classification_report(
truth, predicted, output_dict=True, target_names=labels, digits=3
)
accuracy = accuracy_score(truth, predicted)
if "comprehensiveness_classification_scores" in instances[0]:
comprehensiveness_scores = [
x["classification_scores"][x["classification"]]
- x["comprehensiveness_classification_scores"][x["classification"]]
for x in instances
]
comprehensiveness_score = np.average(comprehensiveness_scores)
else:
comprehensiveness_score = None
comprehensiveness_scores = None
if "sufficiency_classification_scores" in instances[0]:
sufficiency_scores = [
x["classification_scores"][x["classification"]]
- x["sufficiency_classification_scores"][x["classification"]]
for x in instances
]
sufficiency_score = np.average(sufficiency_scores)
else:
sufficiency_score = None
sufficiency_scores = None
if "comprehensiveness_classification_scores" in instances[0]:
comprehensiveness_entropies = [
entropy(list(x["classification_scores"].values()))
- entropy(list(x["comprehensiveness_classification_scores"].values()))
for x in instances
]
comprehensiveness_entropy = np.average(comprehensiveness_entropies)
comprehensiveness_kl = np.average(
list(
compute_kl(
x["classification_scores"],
x["comprehensiveness_classification_scores"],
)
for x in instances
)
)
else:
comprehensiveness_entropies = None
comprehensiveness_kl = None
comprehensiveness_entropy = None
if "sufficiency_classification_scores" in instances[0]:
sufficiency_entropies = [
entropy(list(x["classification_scores"].values()))
- entropy(list(x["sufficiency_classification_scores"].values()))
for x in instances
]
sufficiency_entropy = np.average(sufficiency_entropies)
sufficiency_kl = np.average(
list(
compute_kl(
x["classification_scores"], x["sufficiency_classification_scores"]
)
for x in instances
)
)
else:
sufficiency_entropies = None
sufficiency_kl = None
sufficiency_entropy = None
if "thresholded_scores" in instances[0]:
(
aopc_thresholds,
aopc_comprehensiveness_score,
aopc_comprehensiveness_points,
aopc_sufficiency_score,
aopc_sufficiency_points,
) = compute_aopc_scores(instances, aopc_thresholds)
else:
(
aopc_thresholds,
aopc_comprehensiveness_score,
aopc_comprehensiveness_points,
aopc_sufficiency_score,
aopc_sufficiency_points,
) = (None, None, None, None, None)
if "tokens_to_flip" in instances[0]:
token_percentages = []
for ann in annotations:
# in practice, this is of size 1 for everything except e-snli
docids = set(ev.docid for ev in chain.from_iterable(ann.evidences))
inst = key_to_instances[ann.annotation_id]
tokens = inst["tokens_to_flip"]
doc_lengths = sum(len(docs[d]) for d in docids)
token_percentages.append(tokens / doc_lengths)
token_percentages = np.average(token_percentages)
else:
token_percentages = None
return {
"accuracy": accuracy,
"prf": classification_scores,
"comprehensiveness": comprehensiveness_score,
"sufficiency": sufficiency_score,
"comprehensiveness_entropy": comprehensiveness_entropy,
"comprehensiveness_kl": comprehensiveness_kl,
"sufficiency_entropy": sufficiency_entropy,
"sufficiency_kl": sufficiency_kl,
"aopc_thresholds": aopc_thresholds,
"comprehensiveness_aopc": aopc_comprehensiveness_score,
"comprehensiveness_aopc_points": aopc_comprehensiveness_points,
"sufficiency_aopc": aopc_sufficiency_score,
"sufficiency_aopc_points": aopc_sufficiency_points,
}
def verify_instance(instance: dict, docs: Dict[str, list], thresholds: Set[float]):
error = False
docids = []
# verify the internal structure of these instances is correct:
# * hard predictions are present
# * start and end tokens are valid
# * soft rationale predictions, if present, must have the same document length
for rat in instance["rationales"]:
docid = rat["docid"]
if docid not in docid:
error = True
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, docid={docid} could not be found as a preprocessed document! Gave up on additional processing.'
)
continue
doc_length = len(docs[docid])
for h1 in rat.get("hard_rationale_predictions", []):
# verify that each token is valid
# verify that no annotations overlap
for h2 in rat.get("hard_rationale_predictions", []):
if h1 == h2:
continue
if (
len(
set(range(h1["start_token"], h1["end_token"]))
& set(range(h2["start_token"], h2["end_token"]))
)
> 0
):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, docid={docid} {h1} and {h2} overlap!'
)
error = True
if h1["start_token"] > doc_length:
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, docid={docid} received an impossible tokenspan: {h1} for a document of length {doc_length}'
)
error = True
if h1["end_token"] > doc_length:
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, docid={docid} received an impossible tokenspan: {h1} for a document of length {doc_length}'
)
error = True
# length check for soft rationale
# note that either flattened_documents or sentence-broken documents must be passed in depending on result
soft_rationale_predictions = rat.get("soft_rationale_predictions", [])
if (
len(soft_rationale_predictions) > 0
and len(soft_rationale_predictions) != doc_length
):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, docid={docid} expected classifications for {doc_length} tokens but have them for {len(soft_rationale_predictions)} tokens instead!'
)
error = True
# count that one appears per-document
docids = Counter(docids)
for docid, count in docids.items():
if count > 1:
error = True
logging.info(
'Error! For instance annotation={instance["annotation_id"]}, docid={docid} appear {count} times, may only appear once!'
)
classification = instance.get("classification", "")
if not isinstance(classification, str):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, classification field {classification} is not a string!'
)
error = True
classification_scores = instance.get("classification_scores", dict())
if not isinstance(classification_scores, dict):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, classification_scores field {classification_scores} is not a dict!'
)
error = True
comprehensiveness_classification_scores = instance.get(
"comprehensiveness_classification_scores", dict()
)
if not isinstance(comprehensiveness_classification_scores, dict):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, comprehensiveness_classification_scores field {comprehensiveness_classification_scores} is not a dict!'
)
error = True
sufficiency_classification_scores = instance.get(
"sufficiency_classification_scores", dict()
)
if not isinstance(sufficiency_classification_scores, dict):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, sufficiency_classification_scores field {sufficiency_classification_scores} is not a dict!'
)
error = True
if ("classification" in instance) != ("classification_scores" in instance):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, when providing a classification, you must also provide classification scores!'
)
error = True
if ("comprehensiveness_classification_scores" in instance) and not (
"classification" in instance
):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, when providing a classification, you must also provide a comprehensiveness_classification_score'
)
error = True
if ("sufficiency_classification_scores" in instance) and not (
"classification_scores" in instance
):
logging.info(
f'Error! For instance annotation={instance["annotation_id"]}, when providing a sufficiency_classification_score, you must also provide a classification score!'
)
error = True
if "thresholded_scores" in instance:
instance_thresholds = set(
x["threshold"] for x in instance["thresholded_scores"]
)
if instance_thresholds != thresholds:
error = True
logging.info(
'Error: {instance["thresholded_scores"]} has thresholds that differ from previous thresholds: {thresholds}'
)
if (
"comprehensiveness_classification_scores" not in instance
or "sufficiency_classification_scores" not in instance
or "classification" not in instance
or "classification_scores" not in instance
):
error = True
logging.info(
"Error: {instance} must have comprehensiveness_classification_scores, sufficiency_classification_scores, classification, and classification_scores defined when including thresholded scores"
)
if not all(
"sufficiency_classification_scores" in x
for x in instance["thresholded_scores"]
):
error = True
logging.info(
"Error: {instance} must have sufficiency_classification_scores for every threshold"
)
if not all(
"comprehensiveness_classification_scores" in x
for x in instance["thresholded_scores"]
):
error = True
logging.info(
"Error: {instance} must have comprehensiveness_classification_scores for every threshold"
)
return error
def verify_instances(instances: List[dict], docs: Dict[str, list]):
annotation_ids = list(x["annotation_id"] for x in instances)
key_counter = Counter(annotation_ids)
multi_occurrence_annotation_ids = list(
filter(lambda kv: kv[1] > 1, key_counter.items())
)
error = False
if len(multi_occurrence_annotation_ids) > 0:
error = True
logging.info(
f"Error in instances: {len(multi_occurrence_annotation_ids)} appear multiple times in the annotations file: {multi_occurrence_annotation_ids}"
)
failed_validation = set()
instances_with_classification = list()
instances_with_soft_rationale_predictions = list()
instances_with_soft_sentence_predictions = list()
instances_with_comprehensiveness_classifications = list()
instances_with_sufficiency_classifications = list()
instances_with_thresholded_scores = list()
if "thresholded_scores" in instances[0]:
thresholds = set(x["threshold"] for x in instances[0]["thresholded_scores"])
else:
thresholds = None
for instance in instances:
instance_error = verify_instance(instance, docs, thresholds)
if instance_error:
error = True
failed_validation.add(instance["annotation_id"])
if instance.get("classification", None) != None:
instances_with_classification.append(instance)
if instance.get("comprehensiveness_classification_scores", None) != None:
instances_with_comprehensiveness_classifications.append(instance)
if instance.get("sufficiency_classification_scores", None) != None:
instances_with_sufficiency_classifications.append(instance)
has_soft_rationales = []
has_soft_sentences = []
for rat in instance["rationales"]:
if rat.get("soft_rationale_predictions", None) != None:
has_soft_rationales.append(rat)
if rat.get("soft_sentence_predictions", None) != None:
has_soft_sentences.append(rat)
if len(has_soft_rationales) > 0:
instances_with_soft_rationale_predictions.append(instance)
if len(has_soft_rationales) != len(instance["rationales"]):
error = True
logging.info(
f'Error: instance {instance["annotation"]} has soft rationales for some but not all reported documents!'
)
if len(has_soft_sentences) > 0:
instances_with_soft_sentence_predictions.append(instance)
if len(has_soft_sentences) != len(instance["rationales"]):
error = True
logging.info(
f'Error: instance {instance["annotation"]} has soft sentences for some but not all reported documents!'
)
if "thresholded_scores" in instance:
instances_with_thresholded_scores.append(instance)
logging.info(
f"Error in instances: {len(failed_validation)} instances fail validation: {failed_validation}"
)
if len(instances_with_classification) != 0 and len(
instances_with_classification
) != len(instances):
logging.info(
f"Either all {len(instances)} must have a classification or none may, instead {len(instances_with_classification)} do!"
)
error = True
if len(instances_with_soft_sentence_predictions) != 0 and len(
instances_with_soft_sentence_predictions
) != len(instances):
logging.info(
f"Either all {len(instances)} must have a sentence prediction or none may, instead {len(instances_with_soft_sentence_predictions)} do!"
)
error = True
if len(instances_with_soft_rationale_predictions) != 0 and len(
instances_with_soft_rationale_predictions
) != len(instances):
logging.info(
f"Either all {len(instances)} must have a soft rationale prediction or none may, instead {len(instances_with_soft_rationale_predictions)} do!"
)
error = True
if len(instances_with_comprehensiveness_classifications) != 0 and len(
instances_with_comprehensiveness_classifications
) != len(instances):
error = True
logging.info(
f"Either all {len(instances)} must have a comprehensiveness classification or none may, instead {len(instances_with_comprehensiveness_classifications)} do!"
)
if len(instances_with_sufficiency_classifications) != 0 and len(
instances_with_sufficiency_classifications
) != len(instances):
error = True
logging.info(
f"Either all {len(instances)} must have a sufficiency classification or none may, instead {len(instances_with_sufficiency_classifications)} do!"
)
if len(instances_with_thresholded_scores) != 0 and len(
instances_with_thresholded_scores
) != len(instances):
error = True
logging.info(
f"Either all {len(instances)} must have thresholded scores or none may, instead {len(instances_with_thresholded_scores)} do!"
)
if error:
raise ValueError(
"Some instances are invalid, please fix your formatting and try again"
)
def _has_hard_predictions(results: List[dict]) -> bool:
# assumes that we have run "verification" over the inputs
return (
"rationales" in results[0]
and len(results[0]["rationales"]) > 0
and "hard_rationale_predictions" in results[0]["rationales"][0]
and results[0]["rationales"][0]["hard_rationale_predictions"] is not None
and len(results[0]["rationales"][0]["hard_rationale_predictions"]) > 0
)
def _has_soft_predictions(results: List[dict]) -> bool:
# assumes that we have run "verification" over the inputs
return (
"rationales" in results[0]
and len(results[0]["rationales"]) > 0
and "soft_rationale_predictions" in results[0]["rationales"][0]
and results[0]["rationales"][0]["soft_rationale_predictions"] is not None
)
def _has_soft_sentence_predictions(results: List[dict]) -> bool:
# assumes that we have run "verification" over the inputs
return (
"rationales" in results[0]
and len(results[0]["rationales"]) > 0
and "soft_sentence_predictions" in results[0]["rationales"][0]
and results[0]["rationales"][0]["soft_sentence_predictions"] is not None
)
def _has_classifications(results: List[dict]) -> bool:
# assumes that we have run "verification" over the inputs
return "classification" in results[0] and results[0]["classification"] is not None
def main():
parser = argparse.ArgumentParser(
description="""Computes rationale and final class classification scores""",
formatter_class=argparse.RawTextHelpFormatter,
)
parser.add_argument(
"--data_dir",
dest="data_dir",
required=True,
help="Which directory contains a {train,val,test}.jsonl file?",
)
parser.add_argument(
"--split",
dest="split",
required=True,
help="Which of {train,val,test} are we scoring on?",
)
parser.add_argument(
"--strict",
dest="strict",
required=False,
action="store_true",
default=False,
help="Do we perform strict scoring?",
)
parser.add_argument(
"--results",
dest="results",
required=True,
help="""Results File
Contents are expected to be jsonl of:
{
"annotation_id": str, required
# these classifications *must not* overlap
"rationales": List[
{
"docid": str, required
"hard_rationale_predictions": List[{
"start_token": int, inclusive, required
"end_token": int, exclusive, required
}], optional,
# token level classifications, a value must be provided per-token
# in an ideal world, these correspond to the hard-decoding above.
"soft_rationale_predictions": List[float], optional.
# sentence level classifications, a value must be provided for every
# sentence in each document, or not at all
"soft_sentence_predictions": List[float], optional.
}
],
# the classification the model made for the overall classification task
"classification": str, optional
# A probability distribution output by the model. We require this to be normalized.
"classification_scores": Dict[str, float], optional
# The next two fields are measures for how faithful your model is (the
# rationales it predicts are in some sense causal of the prediction), and
# how sufficient they are. We approximate a measure for comprehensiveness by
# asking that you remove the top k%% of tokens from your documents,
# running your models again, and reporting the score distribution in the
# "comprehensiveness_classification_scores" field.
# We approximate a measure of sufficiency by asking exactly the converse
# - that you provide model distributions on the removed k%% tokens.
# 'k' is determined by human rationales, and is documented in our paper.
# You should determine which of these tokens to remove based on some kind
# of information about your model: gradient based, attention based, other
# interpretability measures, etc.
# scores per class having removed k%% of the data, where k is determined by human comprehensive rationales
"comprehensiveness_classification_scores": Dict[str, float], optional
# scores per class having access to only k%% of the data, where k is determined by human comprehensive rationales
"sufficiency_classification_scores": Dict[str, float], optional
# the number of tokens required to flip the prediction - see "Is Attention Interpretable" by Serrano and Smith.
"tokens_to_flip": int, optional
"thresholded_scores": List[{
"threshold": float, required,
"comprehensiveness_classification_scores": like "classification_scores"
"sufficiency_classification_scores": like "classification_scores"
}], optional. if present, then "classification" and "classification_scores" must be present
}
When providing one of the optional fields, it must be provided for *every* instance.
The classification, classification_score, and comprehensiveness_classification_scores
must together be present for every instance or absent for every instance.
""",
)
parser.add_argument(
"--iou_thresholds",
dest="iou_thresholds",
required=False,
nargs="+",
type=float,
default=[0.5],
help="""Thresholds for IOU scoring.
These are used for "soft" or partial match scoring of rationale spans.
A span is considered a match if the size of the intersection of the prediction
and the annotation, divided by the union of the two spans, is larger than
the IOU threshold. This score can be computed for arbitrary thresholds.
""",
)
parser.add_argument(
"--score_file",
dest="score_file",
required=False,
default=None,
help="Where to write results?",
)
parser.add_argument(
"--aopc_thresholds",
nargs="+",
required=False,
type=float,
default=[0.01, 0.05, 0.1, 0.2, 0.5],
help="Thresholds for AOPC Thresholds",
)
args = parser.parse_args()
results = load_jsonl(args.results)
docids = set(
chain.from_iterable(
[rat["docid"] for rat in res["rationales"]] for res in results
)
)
docs = load_flattened_documents(args.data_dir, docids)
verify_instances(results, docs)
# load truth
annotations = annotations_from_jsonl(
os.path.join(args.data_dir, args.split + ".jsonl")
)
docids |= set(
chain.from_iterable(
(ev.docid for ev in chain.from_iterable(ann.evidences))
for ann in annotations
)
)
has_final_predictions = _has_classifications(results)
scores = dict()
if args.strict:
if not args.iou_thresholds:
raise ValueError(
"--iou_thresholds must be provided when running strict scoring"
)
if not has_final_predictions:
raise ValueError(
"We must have a 'classification', 'classification_score', and 'comprehensiveness_classification_score' field in order to perform scoring!"
)
# TODO think about offering a sentence level version of these scores.
if _has_hard_predictions(results):
truth = list(
chain.from_iterable(Rationale.from_annotation(ann) for ann in annotations)
)
pred = list(
chain.from_iterable(Rationale.from_instance(inst) for inst in results)
)
if args.iou_thresholds is not None:
iou_scores = partial_match_score(truth, pred, args.iou_thresholds)
scores["iou_scores"] = iou_scores
# NER style scoring
rationale_level_prf = score_hard_rationale_predictions(truth, pred)
scores["rationale_prf"] = rationale_level_prf
token_level_truth = list(
chain.from_iterable(rat.to_token_level() for rat in truth)
)
token_level_pred = list(
chain.from_iterable(rat.to_token_level() for rat in pred)
)
token_level_prf = score_hard_rationale_predictions(
token_level_truth, token_level_pred
)
scores["token_prf"] = token_level_prf
else:
logging.info("No hard predictions detected, skipping rationale scoring")
if _has_soft_predictions(results):
flattened_documents = load_flattened_documents(args.data_dir, docids)
paired_scoring = PositionScoredDocument.from_results(
results, annotations, flattened_documents, use_tokens=True
)
token_scores = score_soft_tokens(paired_scoring)
scores["token_soft_metrics"] = token_scores
else:
logging.info("No soft predictions detected, skipping rationale scoring")
if _has_soft_sentence_predictions(results):
documents = load_documents(args.data_dir, docids)
paired_scoring = PositionScoredDocument.from_results(
results, annotations, documents, use_tokens=False
)
sentence_scores = score_soft_tokens(paired_scoring)
scores["sentence_soft_metrics"] = sentence_scores
else:
logging.info(
"No sentence level predictions detected, skipping sentence-level diagnostic"
)
if has_final_predictions:
flattened_documents = load_flattened_documents(args.data_dir, docids)
class_results = score_classifications(
results, annotations, flattened_documents, args.aopc_thresholds
)
scores["classification_scores"] = class_results
else:
logging.info("No classification scores detected, skipping classification")
pprint.pprint(scores)
if args.score_file:
with open(args.score_file, "w") as of:
json.dump(scores, of, indent=4, sort_keys=True)
if __name__ == "__main__":
main()
|