Terje Kristensen

A model for dynamic content based e-learning systems

In this paper we introduce a (mathematical) model for Dynamic Content Based e-learning systems (DCM). The model is based on 3 types of maps: Knowledge Map, Learning Map and Student Map. Each map corresponds, respectively, to a model of the knowledge space, the learning process and students participating in the learning process. After introducing the model we show how the model can be used in mathematical education on a case from arithmetic. We also discuss how the next version of the DCM system may evolve, based on the given model.

A neural network approach to hyphenating Norwegian

This paper presents a method to automatically hyphenate Norwegian by a backpropagation neural network. The original database consists of about 67000 Norwegian words and is developed by the Norwegian Computing Centre for the Humanities at the University of Bergen. The experiments have been done on a PC-platform and on only a sample of words from the database. The hyphenation program has been developed in Cortex-Pro, a general purpose neural network simulation environment.

Comparison of neural network based fingerprint classification techniques

The primary task of this work is to compare classification techniques to decrease the matching time in fingerprint identification. For classification, four different artificial neural networks are tested, as well as a non-linear support vector machine. All classifiers are compared and discussed to find the most suitable one. Automatic fingerprint identification systems (AFIS) are today widely used, but for use in embedded systems with less computational power, it is necessary to create less time-consuming systems. The classifiers splits a fingerprint database into four different subclasses. A multi-layer perceptron network using a backpropagation algorithm has shown to suit this problem best, outperforming both BAM, Hopfleld, Kohonen and just barely SVM, with a correct classification rate of 88.8%. This classification decreases the average matching time with a factor of 3.7.

Two regimes of computer hyphenation - a comparison

This paper presents two different regimes to automatically hyphenate Norwegian text. One method is based on a backpropagation neural network, while the other uses the T/sub E/X algorithm. The two approaches are described and compared. The database consists of about 40,000 Norwegian words.

Edge detection in a lateral inhibition network

The paper proposes a method for edge detection based upon a lateral inhibition neural network. Two types of one-dimensional input patterns, a bar and Mach bands are studied. The model is then generalised to two dimensions to show how to extract the boundary of a two-dimensional object. Finally, the method is used to extract different contour lines of a real image. All the models have been developed in a general purpose neural network environment and simulated on a PC.

Classification of eukaryotic and prokaryotic cells by a backpropagation network

In this paper we show how a Backpropagation neural network is used to classify between eukaryotic and prokaryotic cells. The classification is based on their DNA (Deoxyribonuclei) sequences which are obtained from different databases available on the Internet. The sequences are first preprocessed using a sliding window technique to obtain sub-sequence frequencies, and then normalised to make them comparable.

Two neural network paradigms of phoneme transcription - a comparison

In this work a multi layer perceptron (MLP) and a counterpropagation (CP) network for transcription of Norwegian text is presented. Both paradigms are described and compared. The corpus consists of about 50,000 Norwegian words and their transcriptions, developed by the Norwegian Telecom Research Centre. The transcription scheme used is Sampa for Norwegian. The performance for each network has been tested on about 10,000 unknown Norwegian words.

Design and Development of a Multi-Agent E-Learning System

In this paper the authors present the design and software development of an E-learning system based on a multi-agent MAS architecture. The multi-agent architecture is established on the client-server model. The MAS architecture is combined with the Dynamic Content Manager DCM model of E-learning developed at Bergen University College, Norway. The authors first present the quality requirements of the system before they describe the architectural decisions taken. They then evaluate and discuss the benefits of using a multi-agent architecture. Finally, the MAS architecture is compared with a pure service-oriented architecture SOA to observe that a MAS architecture has a lot of the same qualities as this architecture, in addition to some new ones.

Dynamic Content Manager --- A New Conceptual Model for E-Learning

In this paper a conceptual model for e-learning, which uses elements from learning Objects and Concepts Maps, is introduced. The learning material is divided into atomic units and organized in graphs called Knowledge Map, Learning Map and Student Map. Such a structure provides an easy-to-use navigation interface for existing learning material. Any course content created is stored in a repository for future reference. The model is used to structure a course in geometry for post-graduate teacher students. In teacher education one has to account for the ability to transfer knowledge to students, in addition to the assimilation of knowledge. The established model is therefore discussed in a learning or didactical context.

Transcription of text by incremental Support Vector Machine

This paper deals with an on-going work aimed at developing a Suport Vector Machine (SVM) based incremental learning algorithm in the domain of text-based phoneme transcription for Norwegian language. The motivation for this is to reduce the long computation time witnessed in the batch-learning scenario. A standard SVM algorithm is modified in such a way that it incorporates the new data as it becomes available in time. The transcription scheme used is SAMPA for Norwegian. We conclude the paper with a discussion on further research in this direction.