Alfarrábio: Adding value to an heterogeneous site collection

TheWeb is probably the largest collection of digital resources ever built, and it maintains it’s continuous growth. But this large collection is not a digital library, according to the most accepted sense, since it lacks some fundamental mechanisms. While it is possible to create and maintain a structured site, with powerful cataloging mechanisms and information retrieval, in a way that can be considered as a digital library, the task of grouping several heterogeneous sites with the same purpose is a more complicated problem, and several approaches can be taken. In this article, we report our effort to develop information extraction, cataloging and browsing mechanisms to a set of heterogeneous cultural sites, to be regarded, from the user’s point of view, as a single digital library. Our case study is based on the Alfarrábio project, a cultural cooperative to maintain and support the publication of cultural resources on the Web, both by individuals and non-profit organizations. The cataloguing task of every relevant document (HTML pages are just specific instances of documents) was one of the main concerns of the project; it is supported by automatic tools. We build a semantic net of all related terms (according to predefined and user’s defined relations) crossing all Alfarrábio resources: whole sites, HTML pages, images, travel maps, sound files, XML documents, relational data, etc. This is quite different from the approaches taken by the so-called portals, where information within different channels is not crossed, and resources (and whole sites) are only cataloged in broader terms. Due to the lack of human resources, we developed a set of automatic cataloging tools and dynamically generated browsing tools. The participation of each site creator in the cataloguing of his own information is supported and encouraged. In this article we present and explain the developed tools for information extraction, for cataloging maintenance, and for conceptual browsing.