Surrogater : A Simple Yet Efficient Document Condensation System

This paper describes Surrogater, a simple yet efficient document condensation system that is applicable for commercial use. The system consists of two components, a preprocessing component for automatic generation of key terms for a predefined topic, and a condensation component that produces the condensed versions of on-line documents. To evaluate its performance, we compared Surrogater with five other summarization technologies, including Searchable LEAD, a commercial product. Twenty topics across four domains were evaluated, totalling . 30 documents and 1800 summaries. A two-way ANOVA test suggested that Surrogater performed at least equally well, if not better, compared to other commercial or nearly commercial products. More significantly, the empirical comparison did not show any dramatic differences in performance between Surrogater and Searchable LEAD as reported in an earlier study (Brandow et al. 1995)