ZOOLOGICAL IDENTIFICATION EXPERT SYSTEM

Abstract

Zoological identification is the process, or result, of checking various characters of a specimen in order to assign it to the correct taxa of a previously established classification. Authoritative identification requires considerable experience, a large literature library and a comprehensive research collection. In light of such scarcity, we introduce an object-oriented expert system for this area of identification. In the zoological classification system, similar species are grouped into a genus, genera into a family, families into an order, orders into a class, and classes into a phylum. Thus, the identification knowledge base is represented as a deep knowledge structure which closely models the classification hierarchy. A hybrid is composed of diverse features, or character abstractions, of a taxon's immediate descendants. Hybrids correspond to each taxon of the classification hierarchy, and form the identification structure. Here, species are represented as objects with character properties that distinguish one species object from another. As hybrids are created at each progressive level of the hierarchy, a careful selection of properties can be propagated upwards to form character abstractions which are useful in guiding the identification search at that higher level. To distinguish objects with decreasing detail as one transcends the hierarchy, more general character properties are introduced. Once the object base is created, the identification search is guided by a heuristic function which strengthens the ancestral links of objects whose character properties match the specimen's attributes. Simultaneously, a second heuristic function highlights the more relevant properties likely to describe the specimen. Using this approach, the system is able to identify new organisms as they are added to the knowledge base. It can also help the user to infer probable taxa of unknown organism. Our user interface allows both user and system to share the reasoning and decision-making tasks proportionately to the amount of control the user wants. The user's model simulates how a person consults and learns from an expert in identifying an unfamiliar specimen. Hence, our identification system is designed to play an advisory role with the flexibility needed to support users with varying levels of expertise.