drop.py 9.99 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
11
12
13
14
"""
DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs
https://aclanthology.org/attachments/N19-1246.Supplementary.pdf

DROP is a QA dataset which tests comprehensive understanding of paragraphs. In 
this crowdsourced, adversarially-created, 96k question-answering benchmark, a 
system must resolve multiple references in a question, map them onto a paragraph,
and perform discrete operations over them (such as addition, counting, or sorting).

Homepage: https://allenai.org/data/drop

Acknowledgement: This implementation is based on the official evaluation for `DROP`:
https://github.com/allenai/allennlp-reading-comprehension/blob/master/allennlp_rc/eval/drop_eval.py
"""
Jonathan Tow's avatar
Jonathan Tow committed
15
import inspect
Jon Tow's avatar
Jon Tow committed
16
17
import numpy as np
import re
18
import string
Jonathan Tow's avatar
Jonathan Tow committed
19
import lm_eval.datasets.drop.drop
Jon Tow's avatar
Jon Tow committed
20
from scipy.optimize import linear_sum_assignment
21
from lm_eval.base import Task, rf
Jon Tow's avatar
Jon Tow committed
22
23
from lm_eval.metrics import mean

24

Jonathan Tow's avatar
Jonathan Tow committed
25
_CITATION = """
26
27
28
29
30
31
32
33
34
35
36
@misc{dua2019drop,
    title={DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs}, 
    author={Dheeru Dua and Yizhong Wang and Pradeep Dasigi and Gabriel Stanovsky and Sameer Singh and Matt Gardner},
    year={2019},
    eprint={1903.00161},
    archivePrefix={arXiv},
    primaryClass={cs.CL}
}
"""


silentv0x's avatar
silentv0x committed
37
_ARTICLES = re.compile(r"\b(a|an|the)\b", re.UNICODE)
Anish Thite's avatar
Anish Thite committed
38

39

40
class DROP(Task):
Leo Gao's avatar
Leo Gao committed
41
    VERSION = 1
42
    DATASET_PATH = inspect.getfile(lm_eval.datasets.drop.drop)
Jonathan Tow's avatar
Jonathan Tow committed
43
    DATASET_NAME = None
44

Anish Thite's avatar
Anish Thite committed
45
46
    def has_training_docs(self):
        return True
Jon Tow's avatar
Jon Tow committed
47

Anish Thite's avatar
Anish Thite committed
48
49
50
51
52
53
    def has_validation_docs(self):
        return True

    def has_test_docs(self):
        return False

Jonathan Tow's avatar
Jonathan Tow committed
54
    def training_docs(self):
55
56
57
        if self._training_docs is None:
            self._training_docs = list(map(self._process_doc, self.dataset["train"]))
        return self._training_docs
Jonathan Tow's avatar
Jonathan Tow committed
58
59

    def validation_docs(self):
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
        return map(self._process_doc, self.dataset["validation"])

    def _process_doc(self, doc):
        return {
            "id": doc["query_id"],
            "passage": doc["passage"],
            "question": doc["question"],
            "answers": self.get_answers(doc),
        }

    @classmethod
    def get_answers(cls, qa):
        def _flatten_validated_answers(validated_answers):
            """ Flattens a dict of lists of validated answers.
            {"number": ['1', '8'], ...}
            -> [{"number": ['1'], ...}, {"number": ['8'], ...}]
            """
            vas = []
            for i in range(len(validated_answers["number"])):
                vas.append({
                    "number": validated_answers["number"][i],
                    "date": validated_answers["date"][i],
                    "spans": validated_answers["spans"][i],
                })
            return vas
        answers = []
        answers_set = set()
        candidates = [qa["answer"]] + _flatten_validated_answers(qa["validated_answers"])
        for candidate in candidates:
            answer = cls.parse_answer(candidate)
            if answer in answers_set:
                continue
            answers_set.add(answer)
            answers.append(answer)
        return answers
silentv0x's avatar
silentv0x committed
95
96
97
98
99
100
101
102

    @classmethod
    def parse_answer(cls, answer):
        # NOTE: Everything is returned as a tuple for uniformity and hashability.
        if answer["number"] != "":
            return (str(answer["number"]),)
        if answer["spans"] != []:
            return tuple(answer["spans"])
103
104
105
        return (" ".join([answer["date"]["day"],
                          answer["date"]["month"],
                          answer["date"]["year"]]).strip(),)
Jon Tow's avatar
Jon Tow committed
106

107
108
    def doc_to_text(self, doc):
        return f"Passage: {doc['passage']}\nQuestion: {doc['question']}\nAnswer:"
Jon Tow's avatar
Jon Tow committed
109

110
111
    def doc_to_target(self, doc):
        return " " + ", ".join(doc["answers"][0])
Anish Thite's avatar
Anish Thite committed
112

113
114
115
    def construct_requests(self, doc, ctx):
        """Uses RequestFactory to construct Requests and returns an iterable of
        Requests which will be sent to the LM.
jon-tow's avatar
jon-tow committed
116

117
118
119
120
121
122
123
124
125
        :param doc:
            The document as returned from training_docs, validation_docs, or test_docs.
        :param ctx: str
            The context string, generated by fewshot_context. This includes the natural
            language description, as well as the few shot examples, and the question
            part of the document for `doc`.
        """
        conts = [rf.greedy_until(ctx, ["."])]
        return conts
Jon Tow's avatar
Jon Tow committed
126

Leo Gao's avatar
Leo Gao committed
127
    def process_results(self, doc, results):
Jon Tow's avatar
Jon Tow committed
128
129
        """Take a single document and the LM results and evaluates, returning a
        dict where keys are the names of submetrics and values are the values of
Leo Gao's avatar
Leo Gao committed
130
131
        the metric for that one document

Jon Tow's avatar
Jon Tow committed
132
        :param doc:
Jon Tow's avatar
Jon Tow committed
133
            The document as returned from training_docs, validation_docs, or test_docs.
Leo Gao's avatar
Leo Gao committed
134
135
136
        :param results:
            The results of the requests created in construct_requests.
        """
137
        preds, golds = results, doc["answers"]
silentv0x's avatar
silentv0x committed
138
139
140
141
142
143
144
        max_em = 0
        max_f1 = 0
        for gold_answer in golds:
            exact_match, f1_score = self.get_metrics(preds, gold_answer)
            if gold_answer[0].strip():
                max_em = max(max_em, exact_match)
                max_f1 = max(max_f1, f1_score)
145
146
147
148
        return {
            "em": max_em,
            "f1": max_f1
        }
Jon Tow's avatar
Jon Tow committed
149

silentv0x's avatar
silentv0x committed
150
151
152
153
154
155
156
    def get_metrics(self, predicted, gold):
        """
        Takes a predicted answer and a gold answer (that are both either a string or a list of
        strings), and returns exact match and the DROP F1 metric for the prediction.  If you are
        writing a script for evaluating objects in memory (say, the output of predictions during
        validation, or while training), this is the function you want to call, after using
        :func:`answer_json_to_strings` when reading the gold answer from the released data file.
157
        """
silentv0x's avatar
silentv0x committed
158
159
160
        predicted_bags = self._answer_to_bags(predicted)
        gold_bags = self._answer_to_bags(gold)

161
        if set(predicted_bags[0]) == set(gold_bags[0]) and len(predicted_bags[0]) == len(gold_bags[0]):
silentv0x's avatar
silentv0x committed
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
            exact_match = 1.0
        else:
            exact_match = 0.0

        f1_per_bag = self._align_bags(predicted_bags[1], gold_bags[1])
        f1 = np.mean(f1_per_bag)
        f1 = round(f1, 2)
        return exact_match, f1

    def _answer_to_bags(self, answer):
        if isinstance(answer, (list, tuple)):
            raw_spans = answer
        else:
            raw_spans = [answer]
        normalized_spans = []
        token_bags = []
        for raw_span in raw_spans:
            normalized_span = self._normalize(raw_span)
            normalized_spans.append(normalized_span)
            token_bags.append(set(normalized_span.split()))
        return normalized_spans, token_bags

    def _align_bags(self, predicted, gold):
        """
        Takes gold and predicted answer sets and first finds the optimal 1-1 alignment
        between them and gets maximum metric values over all the answers.
        """
        scores = np.zeros([len(gold), len(predicted)])
        for gold_index, gold_item in enumerate(gold):
            for pred_index, pred_item in enumerate(predicted):
                if self._match_numbers_if_present(gold_item, pred_item):
193
                    scores[gold_index, pred_index] = self._compute_f1(pred_item, gold_item)
Jon Tow's avatar
Jon Tow committed
194
        row_ind, col_ind = linear_sum_assignment(-scores)
silentv0x's avatar
silentv0x committed
195
196

        max_scores = np.zeros([max(len(gold), len(predicted))])
Jon Tow's avatar
Jon Tow committed
197
198
199
200
        for row, column in zip(row_ind, col_ind):
            max_scores[row] = max(max_scores[row], scores[row, column])
        return max_scores

silentv0x's avatar
silentv0x committed
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
    def _compute_f1(self, predicted_bag, gold_bag):
        intersection = len(gold_bag.intersection(predicted_bag))
        if not predicted_bag:
            precision = 1.0
        else:
            precision = intersection / float(len(predicted_bag))
        if not gold_bag:
            recall = 1.0
        else:
            recall = intersection / float(len(gold_bag))
        f1 = (
            (2 * precision * recall) / (precision + recall)
            if not (precision == 0.0 and recall == 0.0)
            else 0.0
        )
Jon Tow's avatar
Jon Tow committed
216
217
        return f1

silentv0x's avatar
silentv0x committed
218
219
220
221
222
223
224
225
226
227
    def _match_numbers_if_present(self, gold_bag, predicted_bag):
        gold_numbers = set()
        predicted_numbers = set()
        for word in gold_bag:
            if self._is_number(word):
                gold_numbers.add(word)
        for word in predicted_bag:
            if self._is_number(word):
                predicted_numbers.add(word)
        if (not gold_numbers) or gold_numbers.intersection(predicted_numbers):
228
229
230
231
232
233
234
235
236
            return True
        return False

    def _is_number(self, text):
        try:
            float(text)
            return True
        except ValueError:
            return False
Jon Tow's avatar
Jon Tow committed
237

silentv0x's avatar
silentv0x committed
238
239
    def _remove_articles(self, text):
        return _ARTICLES.sub(" ", text)
240

silentv0x's avatar
silentv0x committed
241
242
    def _white_space_fix(self, text):
        return " ".join(text.split())
243

silentv0x's avatar
silentv0x committed
244
245
246
247
248
249
    def _remove_punc(self, text):
        exclude = set(string.punctuation)
        if not self._is_number(text):
            return "".join(ch for ch in text if ch not in exclude)
        else:
            return text
250

silentv0x's avatar
silentv0x committed
251
252
    def _fix_number(self, text):
        return str(float(text)) if self._is_number(text) else text
253

silentv0x's avatar
Bug fix  
silentv0x committed
254
    def _tokenize(self, text):
silentv0x's avatar
silentv0x committed
255
        return re.split(" |-", text)
256

silentv0x's avatar
silentv0x committed
257
    def _normalize(self, answer):
258
        tokens = [
259
            self._white_space_fix(self._remove_articles(self._fix_number(self._remove_punc(token.lower()))))
silentv0x's avatar
silentv0x committed
260
            for token in self._tokenize(answer)
261
        ]
Jon Tow's avatar
Fixes  
Jon Tow committed
262
        tokens = [token for token in tokens if token.strip()]
Jon Tow's avatar
Jon Tow committed
263
264
        normalized = " ".join(tokens).strip()
        return normalized
Leo Gao's avatar
Leo Gao committed
265
266
267
268

    def aggregation(self):
        """
        :returns: {str: [float] -> float}
Jon Tow's avatar
Jon Tow committed
269
270
            A dictionary where keys are the names of submetrics and values are
            functions that aggregate a list of metrics
Leo Gao's avatar
Leo Gao committed
271
        """
272
273
274
275
        return {
            "em": mean,
            "f1": mean
        }
Leo Gao's avatar
Leo Gao committed
276
277
278
279

    def higher_is_better(self):
        """
        :returns: {str: bool}
Jon Tow's avatar
Jon Tow committed
280
281
            A dictionary where keys are the names of submetrics and values are
            whether a higher value of the submetric is better
Leo Gao's avatar
Leo Gao committed
282
        """
283
284
285
286
        return {
            "em": True,
            "f1": True
        }