RuDaS (Synthetic Datasets for Rule Learning)

RuDaS (Synthetic Datasets for Ru le Learning) 是一种用于生成包含事实和规则的合成数据集以及用于评估规则学习系统的工具，它克服了现有数据集和适当评估方法的缺点。 RuDaS是高度可参数化的; 例如，可以选择常量，谓词，事实数量，规则的后果 (即完整性)，噪声量 (例如，错误或缺失的事实) 以及规则之间的依赖类型。 此外，RuDaS允许通过计算经典和更新的指标 (包括我们引入的新指标) 来评估规则学习系统的性能。RuDaS (Synthetic Datasets for Rule Learning) is a tool that enables the generation of synthetic datasets comprising facts and rules, and supports the evaluation of rule learning systems, while overcoming the limitations of existing datasets and appropriate evaluation methodologies. RuDaS is highly parameterizable; for example, users can configure constants, predicates, the quantity of facts, the consequent of rules (i.e., completeness), the level of noise (e.g., erroneous or missing facts), and the type of dependencies between rules. Additionally, RuDaS allows for the evaluation of rule learning system performance by calculating both classical and cutting-edge metrics, including novel metrics proposed in this work.

搜集汇总

背景与挑战

背景概述

RuDaS是由IBM Research于2019年发布的合成数据集工具，用于生成包含事实和规则的合成数据，并评估规则学习系统的性能。该工具具有高度可参数化的特点，允许用户选择常量、谓词、事实数量等参数，并通过多种指标评估系统性能。