Modeling the performance of an automated keywording system

摘要

This paper presents a model for predicting the performance of an automated keywording system on each document it keywords. Statistical procedures were investigated in order to protect against incorrect keywording by an automated system designed to behave as a human “expert” indexer keywording documents representing lessons learned from military exercises and operations. The benefit of a statistical model is that it allows the opportunity to manually keyword a document when its predicted performance is low. The specifics of the automated system under consideration are described and various performance measures are considered. The model and variables on which it is based, as well as an illustration of its effectiveness, are presented.