Discovering Internet Resources to enrich a structured Personal information space

The Internet is a tremendous resource where one can find documents to enrich a personal information space. The question is: how can one find relevant documents and how can these be organized into an information space? In this paper, we describe a prototype which aims to provide the user with assistance in these two tasks. Our approach assumes the existence of an initial concept structure set up by the user. This structure may contain only rudimentary descriptions for each concept. The system’s task is to find relevant documents from the Internet and to insert them in the appropriate places in the concept structure. 1. Information Management for Internet Users The amount of information available through the Internet is overwhelming; as a result, most of this information goes unnoticed or gets lost again soon after having been noticed. The problem is not new, it is just being exacerbated by two factors: a sudden growth in the number of information consumers accompanied by acceleration of information production. Information management has thus become a pressing problem: under this heading come several computing disciplines and activities, most notably authoring of information resources, information access and manipulation, as well as information collection, selection and display. The work described here is concerned with the last three activities. A user searching for information on the Internet must describe his information need through a query. Then, upon finding a relevant resource, the user may create a bookmark or even a local copy of this resource. But the collection constituted by all these nuggets of information soon grows into a wasteland for lack of organization instead of a useful, personal information space. An important property of such a space would be structure. This structure could emerge through some automatic processing on the documents in the information space or it could be a manual construct that mirrors the user’s mental model of a given topic. The work described here adopts the latter approach and implements an information space built around a course curriculum. Therefore, we assume that the user has already constructed a concept structure, with rudimentary descriptions of the concepts (e.g. a name) and only a few reference documents attached to them. The task of the system is to seek out potentially relevant documents for each concept and to classify them appropriately. To better explain our approach, let us first analyze the current state of search on the Internet. Figure 1 depicts three users grappling with large amounts of information. The first user spends much time and effort in formulating queries to an information retrieval (IR) service and then in examining the results to determine how relevant they actually are. In the end, he has amassed an amorphous collection of documents. The second user has an assistant that helps with the formulation of queries; he can specify his information need at a more conceptual level; he can also indicate which sites are of particular interest, which sites to avoid, which services should be queried and during what period of time this activity should occur, thus improving network traffic. This user will find it easier to obtain relevant resources; but his information space has no structure and, as a result, the information he has