|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
"""Crowdflower datasets""" |
|
|
|
from __future__ import absolute_import, division, print_function |
|
|
|
import csv |
|
import os |
|
import textwrap |
|
|
|
import six |
|
|
|
import datasets |
|
|
|
|
|
_crowdflower_CITATION = r""" |
|
@inproceedings{van2012designing, |
|
title={Designing a scalable crowdsourcing platform}, |
|
author={Van Pelt, Chris and Sorokin, Alex}, |
|
booktitle={Proceedings of the 2012 ACM SIGMOD International Conference on Management of Data}, |
|
pages={765--766}, |
|
year={2012} |
|
} |
|
""" |
|
|
|
_crowdflower_DESCRIPTION = """ |
|
Collection of crowdflower classification datasets |
|
""" |
|
|
|
DATA_URL = "https://www.dropbox.com/s/ldrcdsv8d9qiwg0/crowdflower.zip?dl=1" |
|
|
|
TASK_TO_LABELS = {'airline-sentiment': ['neutral', 'positive', 'negative'], |
|
'corporate-messaging': ['Information', 'Action', 'Exclude', 'Dialogue'], |
|
'economic-news': ['not sure', 'yes', 'no'], |
|
'political-media-audience': ['constituency', 'national'], |
|
'political-media-bias': ['partisan', 'neutral'], |
|
'political-media-message': ['information', |
|
'support', |
|
'policy', |
|
'constituency', |
|
'personal', |
|
'other', |
|
'media', |
|
'mobilization', |
|
'attack'], |
|
'sentiment_nuclear_power': ['Neutral / author is just sharing information', |
|
'Negative', |
|
'Tweet NOT related to nuclear energy', |
|
'Positive'], |
|
'text_emotion': ['sadness', |
|
'empty', |
|
'relief', |
|
'hate', |
|
'worry', |
|
'enthusiasm', |
|
'happiness', |
|
'neutral', |
|
'love', |
|
'fun', |
|
'anger', |
|
'surprise', |
|
'boredom'], |
|
'tweet_global_warming': ['Yes', 'No']} |
|
|
|
def get_labels(task): |
|
return TASK_TO_LABELS[task] |
|
|
|
class crowdflowerConfig(datasets.BuilderConfig): |
|
"""BuilderConfig for crowdflower.""" |
|
|
|
def __init__( |
|
self, |
|
text_features, |
|
label_classes=None, |
|
process_label=lambda x: x, |
|
**kwargs, |
|
): |
|
"""BuilderConfig for crowdflower. |
|
Args: |
|
text_features: `dict[string, string]`, map from the name of the feature |
|
dict for each text field to the name of the column in the tsv file |
|
label_column: `string`, name of the column in the tsv file corresponding |
|
to the label |
|
data_url: `string`, url to download the zip file from |
|
data_dir: `string`, the path to the folder containing the tsv files in the |
|
downloaded zip |
|
citation: `string`, citation for the data set |
|
url: `string`, url for information about the data set |
|
label_classes: `list[string]`, the list of classes if the label is |
|
categorical. If not provided, then the label will be of type |
|
`datasets.Value('float32')`. |
|
process_label: `Function[string, any]`, function taking in the raw value |
|
of the label and processing it to the form required by the label feature |
|
**kwargs: keyword arguments forwarded to super. |
|
""" |
|
|
|
super(crowdflowerConfig, self).__init__(version=datasets.Version("1.0.0", ""), **kwargs) |
|
|
|
self.text_features = text_features |
|
self.label_column = "label" |
|
self.label_classes = get_labels(self.name) |
|
self.data_url = DATA_URL |
|
self.data_dir = os.path.join("crowdflower", self.name) |
|
self.citation = textwrap.dedent(_crowdflower_CITATION) |
|
def process_label(x): |
|
x=str(x) |
|
if x=="Y": |
|
return "Yes" |
|
if x=="N": |
|
return "No" |
|
return x |
|
self.process_label = process_label |
|
self.description = "" |
|
self.url = "" |
|
|
|
|
|
class crowdflower(datasets.GeneratorBasedBuilder): |
|
|
|
"""The General Language Understanding Evaluation (crowdflower) benchmark.""" |
|
|
|
BUILDER_CONFIG_CLASS = crowdflowerConfig |
|
|
|
BUILDER_CONFIGS = [ |
|
crowdflowerConfig(name="sentiment_nuclear_power", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="tweet_global_warming", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="airline-sentiment", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="corporate-messaging", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="economic-news", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="political-media-audience", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="political-media-bias", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="political-media-message", |
|
text_features={"text": "text"},), |
|
crowdflowerConfig(name="text_emotion", |
|
text_features={"text": "text"},), |
|
] |
|
|
|
def _info(self): |
|
features = {text_feature: datasets.Value("string") for text_feature in six.iterkeys(self.config.text_features)} |
|
if self.config.label_classes: |
|
features["label"] = datasets.features.ClassLabel(names=self.config.label_classes) |
|
else: |
|
features["label"] = datasets.Value("float32") |
|
features["idx"] = datasets.Value("int32") |
|
return datasets.DatasetInfo( |
|
description=_crowdflower_DESCRIPTION, |
|
features=datasets.Features(features), |
|
homepage=self.config.url, |
|
citation=self.config.citation + "\n" + _crowdflower_CITATION, |
|
) |
|
|
|
def _split_generators(self, dl_manager): |
|
dl_dir = dl_manager.download_and_extract(self.config.data_url) |
|
data_dir = os.path.join(dl_dir, self.config.data_dir) |
|
|
|
return [ |
|
datasets.SplitGenerator( |
|
name=datasets.Split.TRAIN, |
|
gen_kwargs={ |
|
"data_file": os.path.join(data_dir or "", "train.tsv"), |
|
"split": "train", |
|
}, |
|
), |
|
] |
|
|
|
def _generate_examples(self, data_file, split): |
|
|
|
process_label = self.config.process_label |
|
label_classes = self.config.label_classes |
|
|
|
with open(data_file, encoding="latin-1") as f: |
|
reader = csv.DictReader(f, delimiter="\t", quoting=csv.QUOTE_NONE) |
|
|
|
for n, row in enumerate(reader): |
|
|
|
example = {feat: row[col] for feat, col in six.iteritems(self.config.text_features)} |
|
example["idx"] = n |
|
|
|
|
|
if self.config.label_column in row: |
|
label = row[self.config.label_column] |
|
label = process_label(label) |
|
if label_classes and label not in label_classes: |
|
continue |
|
example["label"] = label |
|
else: |
|
example["label"] = process_label(-1) |
|
if not example["label"] or not example["text"]: |
|
continue |
|
yield example["idx"], example |
