Automatic Topic Map Generation from Free Text using Linguistic Templates

This work reports on the formulation and use of linguistic templates to identify and extract topic map components (topics, occurrences, associations) from free English text that represents a coherent discourse domain, e.g., travel guides. A topic map is automatically created using these extractions, and the resulting map is evaluated for utility and accuracy. The linguistic templates are refined and extended, based on the initial topic map. This approach hopes to provide a rapid and efficient way of boot-strapping useful topic maps through an iterative cycle of free text extraction, topic map generation and template enhancement.