Meaningful Compound Concept Learning? Wentao Ding
任务目标 对于知识库中的类𝐶，组合本体中的原子谓词， 得到𝐶的子类𝑋，要求𝑋存在有意义的自然语言解 释 “NBA Player” [ a owl:Restriction; rdfs:subClassOf dbo:Person owl:onProperty dbo:league owl:hasValue dbr:National_Basketball_Association ]
相关任务 Ontology Learning from Text Linked Data Mining mostly focuses on the automatic or semi-automatic generation of lightweight taxonomies. Linked Data Mining refers to the process of detecting meaningful patterns in RDF graphs. Concept Learning in Description Logics/OWL is a direction of research that aims at learning schema axioms, such as definitions of classes. Knowledge based Natural Language Interpretation? Adj2ER
可能的路线 取一个有意义的𝐶 枚举可能在𝐶上组合的原子约束𝐴 生成表述 记𝐶与𝐴组合得到𝑋 尝试生成适合𝑋而不适合𝐶的表述 取满足约束的实例 若成功，记𝑋为有意义的 生成表述 取满足约束的实例 基于模式提取实例的可能描述 分析生成当前约束的表述
相关研究 DL-learner Biperpedia Adj2ER