Knowledge-Rich Similarity-Based Classification

摘要

This paper proposes to enhance similarity-based classification with different types of imperfect domain knowledge. We introduce a hierarchy of knowledge types and show how the types can be incorporated into similarity measures. Furthermore, we analyze how properties of the domain theory, such as partialness and vagueness, influence classification accuracy. Experiments in a simple domain suggest that partial knowledge is more useful than vague knowledge. However, for data sets from the UCI Machine Learning Repository, we show that even vague domain knowledge that in isolation performs at chance level can substantially increase classification accuracy when being incorporated into similarity-based classification.