SAOKE

SAOKE的全名是符号辅助的开放知识表达。它是一个手工标记的数据集，包含40,000多个中文句子和对应的事实，形式为SAOKE。它是开放域信息提取任务中最大的公开可用的人类标记数据集。 此数据集具有以下优点: 数据真实开放使用: 遵循OIE系统概念，用原句表达与各类知识兼容的知识: 提供四类知识 (关系、属性、描述、概念) 的统一视图准确表达: 能使用分离的关系短语、缺失信息、隐藏信息等准确地代表事实。SAOKE stands for Symbol-Assisted Open Knowledge Expression. It is a manually annotated dataset containing over 40,000 Chinese sentences and their corresponding facts formatted in the SAOKE structure. It is the largest publicly available human-annotated dataset for open-domain information extraction tasks. This dataset has the following advantages: 1. Authentic and openly usable: It adheres to the Open Information Extraction (OIE) system concept, using original sentences to express knowledge compatible with various types of knowledge; 2. Provides a unified view of four categories of knowledge: relation, attribute, description, and concept; 3. Enables accurate fact representation: It can accurately represent facts using separated relational phrases, missing information, hidden information, and other relevant forms.

搜集汇总

背景与挑战

背景概述

SAOKE（符号辅助的开放知识表达）是一个手工标记的数据集，包含超过40,000个中文句子及对应事实，是开放域信息提取领域最大的公开可用人类标注数据集。它遵循OIE系统概念，提供关系、属性、描述和概念四类知识的统一视图，并能准确表达事实。