This paper presents a technology for information portals that supports multimedia retrieval while focusing at the same time on different ecologies of information provider and content owner environments. This means it addresses cross-media information provision and cross-owner content building. The technology was developed under the FP5 project IRAIA 1 and was applied to two different areas: economic information and cultural information related to puppetry 2. The first area puts more emphasis on multimedia retrieval while the latter one has special requirements in respect to cross-owner content building and content delivery. IRAIA enriches the implementation of multimedia, cross-media, and cross content owner retrieval with a design flexibility that enables an information provider to adapt the delivery channel as well as the content building process to requirements specific to its service. Leading through IRAIA applications in the areas mentioned above we present IRAIA’s system design and architecture, its human-computer interaction modes, and its powerful semantic layer that links the different facets of cross-media delivery and of cross-content building. Context-aware computing, cross-media information provision, cross-owner content building, intecreation. 1 The project IRAIA (IST-1999-10602) was co-funded by the European Commission under the 5th framework programme.

, CONTENT-INDEX, PREFACE. And EXAMPLES OF auxiliary data are a classification system and a thesaurus. An example of the representation of an IR document is reported in Figure 1.This representation is ready to be stored and managed by an operational information retrieval system. This IR document has been extracted from the INSPEC on-line database by the ESA-QUEST information retrieval system. The various different parts of the document have been named according to the terminology previously introduced. This figure gives a clearer understanding of what we mean by structured and unstructured data and the differences existing between these two kinds of data. The information retrieval models which have been proposed for application in the information retrieval area (see References [6] and [7] for a general introduction to the subject) differ considerably from one another, basically due to the different nature and structure of the auxiliary data used in the operations of the IR system bas...

The paper presents an approach to automatically construct hypertext structures out of large document samples for a context-oriented retrieval. It contributes to the design of future infrastructure for large-scale information retrieval systems. As its crucial trait, profiles of usage will furnish the process of semantically mapping. This will be based on automatic text analysis that extract the informative content of a data collection applying functionality provided by a four-level architecture. On top of that, data-mining techniques such as transaction analysis, clustering, and classification will be applied to analyse users' navigation activities in order to reveal contextual information that is implicitly contained in these profiles of usage.

this paper, IR will imply text-based retrieval unless explicitly stated otherwise.