After Ph.D.

My computer science thesis has been to design a methodology of knowledge acquisition for different machine learning and case based reasoning algorithms that may be used to classify and identify biological objects.

• • Knowledge acquisition

In this methodology that is applied to natural sciences, I emphasize the use of a structured questionnaire based upon a descriptive model that follows natural descriptive logics. The meaning of descriptive logics and models is slightly the same as the meaning of description logics and ontologies. But descriptive logics and models are the result of a social, economical, technical and ergonomic analysis of domain usages through use objects, rather than the result of a technological approach based on business objects and formalisms (language and grammar). In fact, even though the formal and technical approach is necessary, the syntax definition is not sufficient in the design process for making useful tools and applications. Designers must be involved in the domain as players, i.e. active participants so as to observe and analyze experts everyday work in order to render a real service on the Internet (e-service). This leads to a semantic analysis of real businesses through social, economical and ergonomic studies before implementing technical solutions. My background knowledge in biology is a fundamental piece of the puzzle in order to be able to understand the expert and end-users' needs from the interior: I am able to feel their hope because I know their work in practice.

• Herbarium

When I came to Reunion island, the idea that I followed with P. Gigord, H. Ralambondrainy and G. Ancel was to apply HyperQuest on the acquisition of plant descriptions coming from the herbarium collection of Pr. T. Cadet at University of Reunion. But the interest of botanical specialists such as J. Figier, D. Strasberg and T. Paillier in the laboratory was not to help them classify and identify species with knowledge bases. They were more concerned about the safeguard of herbarium samples that had been collected by T. Cadet. Then they could have an interest in databases for storing metadata about samples in order to retrieve information of taxa more easily. In fact, the taxonomy of Reunion plants was already acknowledged in the Flora of Mascarene and was no more a subject of research for them. T. Cadet who died in 1987 should have been the expert involved for such enhancement with knowledge bases.

• Systematics of corals

Conversely, as well as for sponges, the taxonomy of corals is not well established, i.e. we don't know if there is a set of 600 species or 1000 species in the Scleractinia Order. For accessing information about a specimen, giving a name to it is the first task for someone who wants to study biodiversity of a geographical area. 3000 specimens from the Mascarene archipelago (Reunion, Mauritius, Rodrigues) have been collected by Pr. G. faure, who is one of renown systematician in the world. So, we decided to make a knowledge base for enhancing this thirty years' field work in 1995 in a project called "Corals of the Mascarene archipelago" in the frame of the enhancement of scientific databases VBD program: "Valorisation des Bases de Données scientifiques" at IREMIA and Multimedia Center.

• IKBS

This VBD program was the thinking frame for reifying the methodology into an Iterative Knowledge Base System (IKBS) that helps specialists to classify their collected specimens, and non-specialists to identify the names of new specimens. Classifying means to define species names of collection samples and put them in a taxonomy according to morphological similarities and discriminating characters. Identifying means to recognize the species names according to a predefined taxonomy. IKBS was conceived and developed with these pragmatic objectives in mind for Computer Assisted Systematics. It is not based on the phylogenetic approach, i.e. the "new Systematics" cladistic approach that study the evolution of taxa within a tree of life. This last scientific classification attempt is useful for advanced research focusing on genetic and biomolecular descriptions (barcoding), but fails to face the need of identifying species for biodiversity monitoring in the fields.

• Telesystematics

Interpretation of observation through descriptions is a subjective task that is tied to personal perceptions and background knowledge. Hopefully, our knowledge producers are experts that are trustworthy because of their experience and scientific reputation in the domain! So, their interpretations are considered to be the best ones. Consequently, they lead the knowledge acquisition phase by defining the descriptive model and entering cases, until they estimate that their knowledge base is made of robust descriptions. Indeed, they can come back on the definition of the descriptive model and then update past cases with instances of new objects, attributes and values. Nevertheless, their descriptions can be noisy. The aim of our iterative process is to become inter-subjective between different persons that constitute a community of practices for gaining a descriptive consensus upon specimens. Telesystematics is the Computer Supported Cooperative Work (CSCW) that we have settled up to give experts the possibility to share and discuss their interpretations of observations by videoconferencing with other experts around the world.