Shall We Trust All Relational Tuples by Open Information Extraction? A Study on Speculation Detection

TL;DR

This paper investigates the problem of detecting speculation at the tuple level in Open Information Extraction, proposing a new task, analyzing relevant data, and introducing a baseline model for this purpose.

Contribution

It formally defines tuple-level speculation detection in OIE, conducts detailed data analysis, and proposes a baseline model for this novel task.

Findings

Analysis of the LSOIE dataset reveals the prevalence of speculative tuples.

The baseline model OIE-Spec provides a starting point for tuple-level speculation detection.

The study highlights the importance of considering speculation in downstream applications.

Abstract

Open Information Extraction (OIE) aims to extract factual relational tuples from open-domain sentences. Downstream tasks use the extracted OIE tuples as facts, without examining the certainty of these facts. However, uncertainty/speculation is a common linguistic phenomenon. Existing studies on speculation detection are defined at sentence level, but even if a sentence is determined to be speculative, not all tuples extracted from it may be speculative. In this paper, we propose to study speculations in OIE and aim to determine whether an extracted tuple is speculative. We formally define the research problem of tuple-level speculation detection and conduct a detailed data analysis on the LSOIE dataset which contains labels for speculative tuples. Lastly, we propose a baseline model OIE-Spec for this new research task.