Dynamic Stochastic Models for Indexes and Thesauri, Identification Clouds, and Information Retrieval and Storage

The first topic of this partial survey paper is that of the growth of adequate lists of key phrase terms for a given field of science or thesauri for such a field. A very rough ‘taking averages’ deterministic analysis predicts monotonic growth with saturation effects. A much more sophisticated realistic stochatic model confirms that. The second, and possibly more important, concept in this paper is that of an identification cloud of a keyphrase (or of other things such as formulas or classification numbers). Very roughly this is (textual) context information that indicates whether a standard keyphrase is present, or, better, should be present, whether it is linguistically recognizable or not (or even totally absent). Identification clouds capture a certain amount of expert information for a given field. Applications include automatic keyphrase assignment and dialogue mediated information retrieval (as discussed in this paper). The problem arises how to generate (semi-)automatically identification clouds and a corresponding enriched weak thesaurus for a given field. A possible (updatable and adaptive) solution is described.