metadata
dataset_info:
- config_name: deduped
features:
- name: text
dtype: string
- name: summary
dtype: string
splits:
- name: train
num_bytes: 25789470056
num_examples: 871910
- name: validation
num_bytes: 2046546693
num_examples: 65388
- name: test
num_bytes: 2056452183
num_examples: 65372
download_size: 12137808791
dataset_size: 29892468932
- config_name: default
features:
- name: text
dtype: string
- name: summary
dtype: string
splits:
- name: train
num_bytes: 38364057181
num_examples: 1207222
- name: validation
num_bytes: 2115665259
num_examples: 67068
- name: test
num_bytes: 2129349762
num_examples: 67072
download_size: 17095133825
dataset_size: 42609072202
- config_name: mini
features:
- name: text
dtype: string
- name: summary
dtype: string
- name: cluster_id
dtype: int64
splits:
- name: train
num_bytes: 2661771759
num_examples: 90000
download_size: 1079431729
dataset_size: 2661771759
configs:
- config_name: deduped
data_files:
- split: train
path: deduped/train-*
- split: validation
path: deduped/validation-*
- split: test
path: deduped/test-*
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
- config_name: mini
data_files:
- split: train
path: mini/train-*
license: cc-by-4.0
task_categories:
- text2text-generation
- summarization
language:
- en
tags:
- legal
source_dataset: NortheasternUniversity/big_patent
BEE-spoke-data/bigpatent-all
Original bigpatent subset "all" converted to hf format, columns renamed, some cleaning applied to "summary" column.
additional configs
- the deduped config, which has the
summarycol deduped via minhash - the mini config, which is a ~1 GB version of the deduped dataset created via a minipile-like clustering+sampling approach