Validating SPONGIA, an expert system for sponge identification

In this article we present the validation of SPONGIA, an expert system to help in the identification of marine sponges. Validation was performed by using data from 82 randomly selected literature descriptions of sponge species. The data gathered by SPONGIA to identify each specimen were obtained from the bibliographical description. The set of cases, in which each case was described by the data gathered by SPONGIA, was presented to five internationally recognised experts in sponge systematics. The identifications generated by SPONGIA were compared with the identifications of these human experts by means of a cluster analysis. The similarity between SPONGIA and human identifications was assessed using four measures: Euclidean distance, City-block distance, Mahalanobis distance and Kappa index. In this article we show that SPONGIA obtains similar quality results to the experts in Porifera systematics. © 1999 Elsevier Science Ltd. All rights reserved.This research has been funded in part by a Generalitat de Catalunya fellowship (FI/91-193) and the CICYT project TESEU (TIC91-0430).Peer Reviewe