Automatic Database Mapping and Translation Methods

摘要

Automatic methods for content-directed translation and information mapping among machine-readable databases are considered. Multicategory decision algorithms, tested against the classification decisions of human experts in a bibliographic document classification task, matched human performance to the 88% level and showed several key parametric sensitivities. The algorithms tested were based on prior state (experiential) descriptor and category usage, and can be configured to provide incremental adaptation or learning. Techniques used emanate from the pattern recognition and artificial intelligence fields. (ERA citation 03:049333)