Georg Walch

A Graph Structured Text Field Index Based on Word Fragments

Abstract An indexing technique for text data based on word fragments is described. In contrast to earlier approaches the fragments are allowed to be overlapping and are linked in a directed graph structure reflecting that many fragments (“Superstrings”) contain other fragments as substrings. This leads to a redundant free set of primary data pointers. By classifying the set of Superstrings belonging to a fragment according to the position of the fragment in the Superstring, one gains a novel possibility of supporting exact match-, partial match-, and masked partial match-retrieval by an index. The search strategies for the various retrieval cases are described.

Uberlegungen zur Datenbank-Kooperation zwischen Server und Workstations

Viele Rechneranwendungen erfordern heute den Einsatz von Arbeitsplatzrechnern (Workstations) in Verbindung mit einem zentralen Universalrechner, der einen Datenbank-Service zur Verfugung stellt. Von besonderer Bedeutung ist dabei die effiziente Kooperation zwischen Datenbank-Server und Workstations. In dieser Arbeit erlautern und begrunden wir das Server-Workstation-Konzept und gehen auf bestehende Ansatze der Kooperation ein. Anhand der Architektur von Non-Standard-Datenbanksystemen klassifizieren wir dann die verschiedenen Moglichkeiten der Kooperation und ordnen sie im Hinblick auf ihre Qualitat ein.

Managing Complex Objects in R2D2

R2D2 — A Relational Robotics Database System with Extensible Data Types — is a joint project of the IBM Heidelberg Scientific Center and the University of Karlsruhe, Computer Science Department. It aims at the design and implementation of a data management system to support engineering (esp. robotics) applications. The management of complex objects in R2D2 is supported by the underlying database system AIM-P. In this paper we describe the AIM-P system in some detail, its data model and database language, and we emphasize the aspects of extensibility by user defined data types and functions.

Ein Dialogsystem zur Methodensuche

Die heute verfugbaren Programmbibliotheken bieten im allgemeinen zu wenig Unterstutzung, um fur ein Problem die Losungsmethode und das zugehorige Programm aufzufinden und unter gegebenenfalls mehreren einsetzbaren auszuwahlen.

Speicherungsstrukturen und Adressierungstechniken Für Komplexe Objekte des NF2-Relationenmodells

Viele Non-Standard-Anwendungen erfordern die Verwaltung komplexer Objekte in Datenbanksystemen. Die Modellierung solcher Objekte wird sich daher oft am NF2-Relationenmodell orientieren. Dabei stellt sich die Frage, wie diese Objekte auf der internen Ebene eines Datenbanksystems zu verwalten sind. Die hier beschriebenen Konzepte, hierarchisches TID-Konzept und Minidirectory Konzept, stellen neue Ansatze zur Speicherung und Adressierung komplexer Objekte dar. Sie sind beide unterschiedliche Erweiterungen des TID-Konzeptes und erfullen die wesentlichen Anforderungen hinsichtlich Speicher-, Retrieval-, Update- und Reorganisationskosten. Eine genauere Analyse und ein Vergleich soll die Eignung dieser Adressierungskonzepte fur unterschiedliche Anwendungen beurteilen helfen.

Information retrieval with APL by adaptive index and user guidance

A system is described applicable for information retrieval and update both in formatted and unformatted files containing non-numerical data. It uses a new reference-string indexing technique which supports partial-match queries and similar-record search. The reference-string index is adapted to data usage or data. Those parts of records which are estimated by the program to be specified very often in queries are included as reference strings and inverted. For data access in the retrieval phase, the specified attribute values and their logical expressions are transformed into a logical expression between reference strings. Primary and secondary (index) data are stored as external files using auxiliary processors for access. The retrieval and update functions are combined with the user-guidance component used to describe data (dictionary), to teach system use, and to assist as a permanently available Help feature. As an internal structure it uses a semantic net to navigate the user from general to more specific information. A new interesting feature is the use of selected reference strings and selected combinations of them as a more detailed data description extracted automatically from the data or even data usage.

A Classification Problem in Medical Radioscintigraphy

While digital image processing in nuclear medicine until recently consisted in a quality improvement by various filter methods, but left the recognition and classification fully to the human inspection and experience, the proposed test is a step towards automatic pattern recognition and classification. Its practical application is still in the beginning stage, but the success with simulated images lets us assume that it will be helpful in the clinical work.