Understanding Natural Language Metadata

Handling everyday tasks such as search, classification and integration is becoming increasingly difficult and sometimes even impossible due to the increasing streams of data available. To overcome such an information overload we need more accurate information processing tools capable of handling big amounts of data. In particular, handling metadata can give us leverage over the data and enable structured processing of data, however, while some of this metadata is in a computer readable format, some of it is manually created in ambiguous natural language. Thus, accessing the semantics of natural language can increase the quality of information processing. We propose a natural language metadata understanding architecture that enables applications such as semantic matching, classification and search based on natural language metadata by providing a translation into a formal language which outperforms the state of the art by 15%.