Lightweight Parsing of Classifications

Understanding metadata written in natural language is a crucial requirement towards the successful automated integration of large scale, language-rich, classifications such as the ones used in digital libraries. In this article we analyze natural language labels used in such classifications by exploring their syntactic structure, and then we show how this structure can be used to detect patterns of language that can be processed by a lightweight parser whose average accuracy is 96.82%. This allows for a deep understanding of natural language metadata semantics. In particular we show how we improve the accuracy of the automatic translation of classifications into lightweight ontologies by almost 18% with respect to the previously used approach. The automatic translation is required by applications such as semantic matching, search and classification algorithms.