LSOIE: A Large-Scale Dataset for Supervised Open Information Extraction

TL;DR

This paper introduces LSOIE, a large-scale dataset for supervised open information extraction, significantly expanding the resources available for training and evaluating OIE systems, and providing benchmarks for future research.

Contribution

The paper presents LSOIE, a new large-scale, diverse dataset for supervised OIE, derived from QA-SRL 2.0, and establishes baseline models and benchmarks.

Findings

LSOIE is 20 times larger than previous datasets.

Benchmark models achieve baseline performance on LSOIE.

Data, models, and code are publicly available for research.

Abstract

Open Information Extraction (OIE) systems seek to compress the factual propositions of a sentence into a series of n-ary tuples. These tuples are useful for downstream tasks in natural language processing like knowledge base creation, textual entailment, and natural language understanding. However, current OIE datasets are limited in both size and diversity. We introduce a new dataset by converting the QA-SRL 2.0 dataset to a large-scale OIE dataset (LSOIE). Our LSOIE dataset is 20 times larger than the next largest human-annotated OIE dataset. We construct and evaluate several benchmark OIE models on LSOIE, providing baselines for future improvements on the task. Our LSOIE data, models, and code are made publicly available