Cyc-Enhanced Machine Classification
msra(2009)
摘要
We describe a framework for linking together a structured ontology, deductive logic, and probability, to solve classification problems. We illustrate this framework with the Whodunit problem: identifying the perpetrator of a crime. Several experiments show that the use of Cyc's ontology and inference abilities substantially improves classification accuracy, both in decision tree classifiers and with Markov Logic Networks.
更多查看译文
关键词
decision tree classifier
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络