Probability estimation is important for the application of probabilistic models as well as for any evaluation in IR. We discuss the interdependencies between parameter estimation and certain properties of probabilistic models: dependence assumptions, binary vs. non-binary features, estimation sample selection. Then we define an optimum estimate for binary features which can be applied to various typical estimation problems in IR. A method for computing this estimate using empirical data is described. Some experiments show the applicability of our method, whereas comparable approaches are partially based on false assumptions or yield biased estimates.

Proceedings of the Fourteenth Annual International ACM SIGIR Conference on Research and Development in Information Retrieval

With the field capabilities of freeWAIS-sf it became difficult to query more than one database in parallel. The set of searchable fields can differ for different databases, they may have heterogeneous schemas. An important concept in database systems is data independence. Based on this concept a unifying view on multiple freeWAIS-sf databases is presented: the aspect of heterogeneity of different databases is hidden from the user. SFgate is a gateway between the World Wide Web and freeWAIS-sf which implements this approach.

WP4 deals with automatic categorisation of web documents that is based on a description oriented approach to document indexing. This deliverable describes further progress with respect to the work done in Deliverable 4.1 as well as an Intermediate Prototype which implements parts of the architecture given in Deliverable 4.1.

We describe the design and implementation of a system for logic-based multimedia retrieval. As high-level logic for retrieval of hypermedia documents, we have developed a probabilistic object-oriented logic (POOL) which supports aggregated objects, different kinds of propositions (terms, classifications and attributes) and even rules as being contained in objects. Based on a probabilistic four-valued logic, POOL uses an implicit open world assumption, allows for closed world assumptions and is able to deal with inconsistent knowledge. POOL programs and queries are translated into probabilistic Datalog programs which can be interpreted by the HySpirit inference engine. For storing the multimedia data, we have developed a new basic IR engine which yields physical data abstraction. The overall architecture and the flexibility of each layer supports logic-based methods for multimedia information retrieval.

In the near future, users will have access to a vast number of digital libraries. For a given information need and limited resources, there is the problem of selecting those libraries which produce an overall optimum answer. This resource discovery problem is additionally complicated by the diversity of the sources, e.g. with respect to media, document formats, indexing methods, database schemas and protocols. Once a set of digital libraries has been selected, the collection fusion problem deals with the problem of merging the answers of these libraries in order to get a high retrieval quality. This paper describes the specific problems and gives an overview on the solutions that have been developed so far.

The automatic categorisation of web documents is becoming crucial for organising the huge amount of information available in the Internet. We are facing a new challenge due to the fact that web documents have a rich structure and are highly heterogeneous. Two ways to respond to this challenge are (1) to use a representation of the content of web documents that captures these two characteristics and (2) to use more effective classifiers. Our categorisation approach is based on a probabilistic description-oriented representation of web documents, and a probabilistic interpretation of the k-nearest neighbour classifier. With the former, we provide an enhanced document representation that incorporates the structural and heterogeneous nature of web documents. With the latter, we provide a theoretical sound justification for the various parameters of k-nearest neighbour classifier. Experimental results show that (1) using an enhanced representation of web documents is crucial for an effective categorisation of web documents, and (2) a theoretical interpretation of the k-nearest neighbour classifier gives us improvement over the standard k-nearest neighbour classifier.

Die Zahl digitaler Bibliotheken im WWW wächst. Literaturrecherchen lassen sich immer öfter ohne den Gang in die reale Bibliothek durchführen. Wir beschreiben den Entwurf eines Agentensystems, welches dem Benutzer / der Benutzerin Funktionalität zur Literaturrecherche auf einer höheren Ebene anbietet, als sie von einzelnen digitalen Bibliotheken erbracht werden kann. Dazu wird die von digitalen Bibliotheken angebotene Funktionalität ausgenutzt und kombiniert; die Basiseigenschaften von Agenten, Adaptivität, Kommunikationsfähigkeit und Autonomie, ermöglichen das auf flexible Weise

Despite the fact that many Digital Libraries (DLs) are available on the Internet, users cannot effectively use them because of inadequate functionality, deficient visualisation and insufficient integration of different DLs. In the framework of this project we develop a user-oriented access system for DLs which overcomes these drawbacks. Based on experiences from the librarian area, higher functions to assist proved search strategies will be implemented. Different DLs will be tightly integrated, so that system-wide search and navigation is possible. The system will be adaptive towards different user wishes, regarding preferences concerning content and system involvement.

Today, peer-to-peer services can comprise a large and growing number of services, e.g. search services or services dealing with heterogeneous schemas in the context of Digital Libraries. For a given task, the system has to determine suitable services and their processing order (execution plan). As peers can join or leave the network spontaneously, static execution plans are not sufficient. This paper proposes a logic-based approach for dynamically computing execution plans: Services are described in the DAML-S language. These descriptions are mapped onto Datalog. Finally, logical rules are applied on the service description facts for determining matching services and finding an optimum execution plan.