language:
- en
license: cc-by-nc-sa-2.0
size_categories:
- 1B<n<10B
task_categories:
- text2text-generation
dataset_info:
features:
- name: sentence1
dtype: string
- name: sentence2
dtype: string
- name: label
dtype: float64
splits:
- name: train
num_bytes: 774819382
num_examples: 6076768
- name: validation
num_bytes: 193544580
num_examples: 1519192
- name: test
num_bytes: 50960096
num_examples: 399788
download_size: 610388435
dataset_size: 1019324058
configs:
- config_name: default
data_files:
- split: train
path: data/train-*
- split: validation
path: data/validation-*
- split: test
path: data/test-*
tags:
- rdf-to-text
📘 Dataset Card: KELM Single-Triple Sample
This is a filtered sample of the original KELM (Knowledge Enhanced Language Model) corpus, created and released by Google Research. The sample contains only the entries with a single triple, offering a more focused subset for lightweight experimentation or simplified use cases such as triple-to-text generation or controlled text generation.
🔍 Key Details:
The full dataset is available in the official KELM repository.
Only entries where the "triples" field has a length of 1 were included in this sample.
📄 License: This dataset is licensed under CC BY-SA 2.0, in accordance with the original release by Google Research. This sample is a redistribution made publicly available for convenience and educational/research purposes on Hugging Face Datasets. No claim of authorship or originality is made.