Annotating the Semantic Web with Natural Language

The Problem: Semantic Web research focuses on imbuing ordinary Web pages with “semantics, ” so that specially enabled software agents can help users better locate, collate, and compare data. Since natural language is the most intuitive information access method for humans, we believe that the Semantic Web should be “natural language enabled. ” However, the application of natural language techniques to the Semantic Web is largely unexplored. Motivation: The vision of the Semantic Web [1] grew out of the recognition that although a wealth of information readily exists today in electronic form, it cannot be easily processed by computers due to a lack of “semantics.” Because computers cannot understand the contents of documents, they are unable to generate new knowledge from existing documents. Because computers cannot related the contents of multiple documents, they are unable to synthesize and compile knowledge from multiple sources. These are two key issues that the Semantic Web is attempting to address. Previous Work: Semantic Web research has been focused on the standardization, structuring, and sharing of on-tologies in a variety of domains. The short-term goal is to transform existing sources (stored as HTML pages, in legacy databases, etc.) into a machine-understandable form (i.e., RDF) consistent with Semantic Web ontologies. Although the general framework of the Semantic Web includes provisions for natural language technology, such techniques have largely been ignored