Datasets:
annotations_creators:
- machine-generated
language_creators:
- found
language:
- en
license:
- mit
multilinguality:
- monolingual
size_categories:
- unknown
source_datasets:
- extended|qa_srl
task_categories:
- text-retrieval
task_ids: []
pretty_name: LSOIE
tags:
- Open Information Extraction
Dataset Card for LSOIE
Table of Contents
- Dataset Description
- Dataset Structure
- Dataset Creation
- Considerations for Using the Data
- Additional Information
Dataset Description
- Homepage: https://github.com/Jacobsolawetz/large-scale-oie
- Repository: https://github.com/Jacobsolawetz/large-scale-oie
- Paper: https://arxiv.org/abs/2101.11177
- Leaderboard: [Needs More Information]
- Point of Contact: [Needs More Information]
Dataset Summary
The Large Scale Open Information Extraction Dataset (LSOIE), is a dataset 20 times larger than the next largest human-annotated Open Information Extraction (OIE) dataset. LSOIE is a built upon the QA-SRL 2.0 dataset by transforming the list of Questions and answers for each predicate to a tuple representing a fact.
Supported Tasks and Leaderboards
Open Information Extraction
Languages
The text in this dataset is english.
Dataset Structure
Data Instances
A datapoint comprises one fact together with the sentence it was extracted from. There can be multiple facts for each Sentence. Each fact is represented by a tuple $(a_0, p, a_1,\dots a_n)$ where $a_0$ is the head entity $p$ is the predicate and $a_1, \dots,a_n$ represent the tail.
Data Fields
- word_ids : sequence of indices (int) representing tokens in a sentence,
- words : a sequence of strings, the tokens in the sentence,
- pred : the predicate of the fact,
- pred_ids : ids of the tokens in the predicate,
- head_pred_id : id of the head token in the predicate,
- sent_id : sentence id,
- run_id : ,
- label : Sequence of tags (BIO) representing the fact, e.g. if the fact is given by $(a_0, p, a_1, \dots, a_n) $
Data Splits
[Needs More Information]
Dataset Creation
Curation Rationale
[Needs More Information]
Source Data
Initial Data Collection and Normalization
[Needs More Information]
Who are the source language producers?
[Needs More Information]
Annotations
Annotation process
[Needs More Information]
Who are the annotators?
[Needs More Information]
Personal and Sensitive Information
[Needs More Information]
Considerations for Using the Data
Social Impact of Dataset
[Needs More Information]
Discussion of Biases
[Needs More Information]
Other Known Limitations
[Needs More Information]
Additional Information
Dataset Curators
[Needs More Information]
Licensing Information
[Needs More Information]
Citation Information
[Needs More Information]