2.3

Processing descriptions for classification and identification

In our applications, all descriptions that are processed are pre-classified : we don't use descriptions for aggregation classification (or categorization). We call a caseor an example the association of a description with an identification from the specialist. For classification purpose, a decision tree is grown using inductive learning from examples. This general knowledge extracted from the examples allows the classes to be characterized (or intentionally defined). For identification purpose, a case based reasoning strategy is used to directly compare the examples in extension (Conruyt et al., 1992).

3.

A brief illustration

A real world application of the ideas presented above has been developped by Pr. C. Lévi, a renowned specialist of Marine Sponges. Using the interactive tools we have designed, he defined a descriptive model concerning the genus Hyalonema (including morphological and histological characters, as well as contextual informations about the specimens). Using a questionnaire automatically built from the descriptive model, he acquired most of the descriptions already published in the literature in a "standard" manner, allowing him to test the different systematical concepts he was using for many years, and to iteratively derive an improved method of description. He presently applies this method in order to propose a new set of the descriptions that he had already published himself, as a way to transmit to other or future specialists of Sponges an optimized package of information about his domain of expertise (Conruyt et al., 1993).

Acknowledgements

Thanks to Prof. E. Diday, Drs J. Lebbe and R. Vignes at INRIA for their valuable advices. And to M. Manago for extensive discussion and emendation of the text.

References

Conruyt N., Manago M., & Le Renard J. (1992) "Modélisation, Formalisation et Analyse d'Objets biologiques en vue de leur identification: application au domaine des éponges marines.", Actes des 3^èmes journées "Symboliques- numériques", Université Paris-IX-Dauphine, 1992.

Conruyt N., Manago M., Le Renard J. & Levi C. (1993) "Une méthode d'acquisition de connaissances pour la classification et l' identification d'objets biologiques.", actes des treizièmes journées sur les systèmes experts et leurs applications, EC2, Avignon, 1993.

Diday E. (1991) "Des objets de l'analyse des données à ceux de l'analyse des connaissances", in "Induction symbolique-numérique à partir de données", vol 1, p. 9-75, Kodratoff Y., Diday E., editions Cepaduès, Toulouse, 1991.

Manago M., Conruyt N. & Le Renard J. (1992) "Acquiring Descriptive Knowledge for Classification and Identification". in: Th. Wetter, K.-D. Althoff, J. Boose, B. Gaines, M. Linster & F. Schmalhofer (eds.), Current Developments in Knowledge Acquisition - EKAW ´92, Springer Verlag.