Unsupervised Ontology-based Semantic Tagging for Knowledge Markup
Paul Buitelaar, Srikanth Ramaka · 2005
A promising approach to automating knowledge markup for the Semantic Web is the application of information extraction technology, which may be used to instantiate classes and their attributes directly from textual data. An important prerequisite for information extraction is the identification and classification of linguistic entities (single words, complex terms, names, etc.) according to concepts in a given ontology. Classification can be handled by standard machine learning approaches, in which concept classifiers are generated by the collection of context models from a training set. Here we describe an unsupervised approach to concept tagging for ontology-based knowledge markup. We discuss the architecture of this system, and our strategy for and results of performance evaluation. 1.