Contributions in Merlin

We report on the treatment of ambiguity in ExtrAns, a system that performs an exhaustive linguistic analysis of UNIX manpages to do answer extraction over them. Disambiguation is performed in two stages. The first stage consists of a set of simple rules that delete some of the wrong interpretations that can be spot with purely syntactic information. The second stage extends the use of Brill and Resnik's algorithm to disambiguate several types of attachment ambiguities. Ambiguities that pass the disambiguation procedures are handled by ExtrAns by displaying the answers to the user with graded selective highlighting.

Stadträte sind sehr stark zeitlich belastet. Neben dem Hauptberuf sind sie ehrenamtlich bis über 40 Stunden pro Woche für die Kommune tätig. Da sie in der Verwaltung kein Büro haben, müssen sie sich abends und am Wochenende zu Hause auf Ihre Sitzungen und Entscheidungen vorbereiten. Zu dieser Zeit sind in der kommunalen Verwaltung keine Mitarbeiter. Das Projekt Computerunterstützung der Parlamentsarbeit (Cuparla) versetzt sie in die Lage, mit Hilfe eines Notebooks Gemeinderatsarbeit zu Hause und an jedem anderen Ort durchzuführen und dabei auf Informationsbestände der Verwaltung, der Fraktion und von Kollegen zuzugreifen. Im Rahmen dieses Projektes wurde eine Software auf der Basis von Domino entwickelt. Cuparla wurde in einem Pilotprojekt mit dem Gemeinderat Stuttgart von Anfang 1996 bis März 1998 erprobt. Es wurden 56 von 57 aktiven Stadträten ausgestattet1. In einem zweiten Feld in Kornwestheim wurden 20 von 26 Stadträten im Herbst 1997 ausgestattet, um den Nutzen auch für kleinere Gemeinden zu evaluieren. Im Folgenden werden zuerst die Ziele des Projekts Cuparla vorgestellt. Danach wird der Bedarf für eine Unterstützung aus der zeitlichen Belastung der Stadträte, ihrer Mobilität und der Informations- und Kooperationsintensität ihrer Arbeit abgeleitet. Sodann wird die Cuparla-Software vorgestellt. Zum Abschluß wird die erweiterte Wirtschaftlichkeit von Cuparla diskutiert.

Unter Telekooperation versteht man die Zusammenarbeit von Personen und Unternehmen, die sich an verschiedenen Standorten befinden. Telekooperation verbindet organisatorische und technische Fragen auf mehreren Ebenen. Individuen benötigen geeignete Werkzeuge, um informell oder in Gruppen besser zusammenarbeiten zu können. Die Werkzeuge führen aber auch zu einer Änderung der individuellen Arbeitsorte (Telearbeit), der Zusammenarbeit in Gruppen (z.B. Ad-hoc-Gruppe) und der Organisation von Unternehmen. In Telekooperationsszenarien werden bestimmte Konfigurationen von Technik und Organisation zusammengeführt.

null

The aim of this work is to conduct research into synergies between Petri-net theory and reverse engineering. The existence of such synergies is not obvious because each sector is based on different assumptions. These differences relate to two modelling paradigms: clustering and folding. Clustering merges neighboured nodes and corresponds to the construction of complex systems from subsystems. It is widely used in software engineering and in practical applications of Petri nets. Foldings only merge transitions with transitions, places with places and arcs with arcs. They group similar functionality. Hence they preserve behaviour, allow the transfer of semantics and provide deep theoretical insights by means of far-reaching connections to other models of concurrency. A folding-based Petri-net algorithm for reverse engineering is introduced. It recovers a coloured net from an unstructured flat Petri net. The two nets are connected by a folding which amounts to a compact specification of the source net. The algorithm is both flexible and scalable, and this work shows how application heuristics can be integrated into it. Its cost is almost linear with respect to the size of the input net, which is remarkable in the field of reverse engineering. Petri nets may serve as an intuitive model of the interplay of the structural, functional and dynamic aspects of a system. Various methods of modelling aspects apart from concurrency by Petri nets represent an innovation. With such a translation, the algorithm may also analyse legacy systems outside the realms of Petri nets. The result is a novel and powerful method of reverse engineering. A specific example shows how a high-level design may be recovered from low-level implementation information. Moreover, the recovered colouring contains a complete specification inclusive of the data model. The reverse engineering part of this work concentrates on folding-based Petri-net methods because they contain new features. On the other hand, clustering-based techniques share many similarities with known methods. For best results, however, clustering and folding should be appropriately combined. The foundations for such combinations are laid down in the first part of this thesis. Many Petri-net classes known from the literature may be grouped into folding-based and clustering-based types. However, no well-defined link exists between them which would allow the strengths of both approaches to be combined, especially for practical applications. Such a link is presented here in the form of an adjunction, which is a strong two-way relationship taken from category theory. It links folding-based and clustering-based categories. It is shown that these categories have properties typical of folding-based and clustering-based Petri-nets respectively. Further compatible adjunctions express the Petri-net dichotomy of structure and behaviour. To the best of the author's knowledge, this basic principle of Petri-net theory has not yet been formulated categorically. For practical applications, it is important that these concepts can be integrated fairly easily with existing Petri-net tools. This will enrich them with the power of a categorical machinery, e.g. with morphisms, universal constructions and the transfer of behaviour. Coloured nets are simply defined as special comma categories, i.e. essentially folding morphisms. The reduction algorithm introduced here is a proof of the practical value of this approach: it is an iteration of couniversal constructions and the reduction itself has couniversal properties.