metadata
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: test
path: data/test-*
dataset_info:
features:
- name: label
dtype:
class_label:
names:
'0': negative
'1': positive
- name: title
dtype: string
- name: content
dtype: string
splits:
- name: train
num_bytes: 163359702
num_examples: 360000
- name: test
num_bytes: 18182813
num_examples: 40000
download_size: 120691417
dataset_size: 181542515
Amazon Polarity 10pct
This is a direct subset of the original Amazon Polarity dataset, downsampled 10pct with a random shuffle
Dataset Summary
For quicker testing on Amazon Polarity. See https://huggingface.co/datasets/amazon_polarity for details and attributions
Source Data
from datasets import ClassLabel, Dataset, DatasetDict, load_dataset
ds_full = load_dataset("amazon_polarity", streaming=True)
ds_train_10_pct = Dataset.from_list(list(ds_full["train"].shuffle(seed=42).take(360_000)))
ds_test_10_pct = Dataset.from_list(list(ds_full["test"].shuffle(seed=42).take(40_000)))
ds_10_pct = DatasetDict({"train": ds_train_10_pct, "test": ds_test_10_pct})
# Need to recreate the class labels
class_label = ClassLabel(num_classes=2, names=["negative", "positive"])
ds_10_pct = ds_10_pct.map(lambda row: {"title": row["title"], "content": row["content"], "label": "negative" if not row["label"] else "positive"})
ds_10_pct = ds_10_pct.cast_column("label", class_label)