Truth in the Digital Library: From Ontological to Hermeneutical Systems [chapter]

Aurélien Bénel, Elöd Egyed-Zsigmond, Yannick Prié, Sylvie Calabretto, Alain Mille, Andréa Iacovella, Jean-Marie Pinon
2001 Lecture Notes in Computer Science  
This paper deals with the conceptual structures which describe document contents in a digital library. Indeed, the underlying question is about the truth of a description: obvious (ontological), by convention (normative) or based on interpretation (hermeneutical). In the first part, we examine the differences between these three points of view and choose the hermeneutical one. Then in the second and third part, we present two "assisted interpretation systems" (AIS) for digital libraries
more » ... sual documents and scholarly publications). Both provide a dynamic annotation framework for readers' augmentations and social interactions. In the fourth part, a few synthetic guidelines are given to design such "assisted interpretation systems" in other digital libraries. * The authors are truly indebted to Phyllis Graham for her help with the English manuscript. One can argue that Ontology and its metaphysical aspects are not to be confused with the ontologies which are proposed by computer scientists. But, as we will see, the simple use of the term "ontology" implies the admission of its a priori understanding of reality. In the first part, we will discuss the postulates and consequences of the ontological approach and introduce the conventional and hermeneutical ones. We firmly support the hermeneutical approach (i.e. based on interpretation) and its interactive and collaborative aspects. In the next two parts, we will present two assisted interpretation systems (AIS) for digital libraries: E-SIA and Porphyry 2000. In the last part, we will propose some general criteria for building hermeneutical digital library systems. Ontology and hermeneutics The particular concern of computer science in "ontologies" comes from its original relationship to reality. On the one hand, computer science derives from modern mathematics which are based on conventionalism. In this approach, there is no concern with the conformity to the real world. Only the internal coherence of a conventionally true ("axiomatic") system matters. On the other hand, computer science as a support for human practices is an applied science and should keep a connection between its formal results and the "real world". The ontological approach dates from Aristotle [2]. By definition, it is the metaphysical study of being or, more pragmatically, the a priori structure of reality. It is based on the utterance of obviously true principles ("true" means "complying with reality"). This way, every deduced proposition complies with reality. Much discussions has concerned Ontology, but nobody has yet managed to get it. We know several "local ontologies". However they are often mutually contradictory and hard to reuse in different applications [6] . There are also many great philosophers, logicians and computer scientists who have tried to find the "top-level ontology" [11]: starting from Aristotle's categories, to Ontolingua [13], and passing through Pierce's categories. But we can't help questioning the ontological approach itself when we read the lengthy catalogue [16] of these works. The "apriorism" of the ontological approach does not correspond to the practice of those who take part in the consensual elaboration of thesauri or domain models. And in fact, the "ontology" concept, in its postmodern meaning, is a mix of a priori truth, conventional truth and consensual truth. It is often referred as a "common ground" for communication within a group [18] or as capturing consensual knowledge as accepted by a group [13] [6] . Indeed experts know that constructing a model is a slow, difficult, collective and regularly repeated process. Even in well defined and formalized domains (medicine, zoology), experts seldom agree on well established concepts [17] . As a matter of fact, the knowledge modeler has had to abandon the ontological "apriorism" in order to continue working. "Apriorism" did not provide many guidelines on how to handle models through their creation, evolution, shared use, learning by a human, fusion, and dependence on practices. On the contrary, these aspects of collaborative construction are well represented in the hermeneutic approach. Traditionally, hermeneutics deals with the production and
doi:10.1007/3-540-44796-2_31 fatcat:tkdjcahr6bbuxocsel3tqb3gva