Vol. 26 · No. 6

This paper outlines a design framework of an intelligent tutoring system (ITS). ITS focuses on a newer and more comprehensive distance learning (DL) process as compared to the established traditional DL programs practiced today. The DL model presented in this paper (CHARLIE) is a high level software based tutorial that has the ability to encompass a wide variety of current DL technologies in a single DL session. CHARLIE's architecture has four components: Control Component (responsible for the interaction between software agents and the operating system); Instructional Component (concerned with the instructional aspects of an ITS session); Text Analysis Component (analyzes the partial syntax and partial semantics of the text in the session); Student Modeling Component (analyzes a student's progress and determines the best model for learning during a session). Each component is serviced by a set of software agents to accomplish its mission. Three additional entities in CHARLIE are two separate databases and an explanation facility. Six agents have been implemented in CHARLIE to create a DL course in C++ programming. Much of CHARLIE remains to be completed which opens many areas for research.

Speech communication interfaces (SCI) are nowadays widely used in several domains. Automated spoken language human-computer interaction can replace human-human interaction if needed. Automatic speech recognition (ASR), a key technology of SCI, has been extensively studied during the past few decades. Most of present systems are based on statistical modeling, both at the acoustic and linguistic levels. Increased attention has been paid to speech recognition in adverse conditions recently, since noise-resistance has become one of the major bottlenecks for practical use of speech recognizers. Although many techniques have been developed, many challenges still have to be overcome before the ultimate goal – creating machines capable of communicating with humans naturally – can be achieved. In this paper we describe the research and development of the first Slovak spoken language dialogue system. The dialogue system is based on the DARPA Communicator architecture. The proposed system consists of the Galaxy hub and telephony, automatic speech recognition, text-to-speech, backend, transport and VoiceXML dialogue management modules. The SCI enables multi-user interaction in the Slovak language. Functionality of the SLDS is demonstrated and tested via two pilot applications, ``Weather forecast for Slovakia'' and ``Timetable of Slovak Railways''. The required information is retrieved from Internet resources in multi-user mode through PSTN, ISDN, GSM and/or VoIP network.

The paper studies a distributed iterative algorithm for optimal scheduling in grid computing. Grid user's requirements are formulated as dimensions in a quality of service problem expressed as a market game played by grid resource agents and grid task agents. User benefits resulting from taking decisions regarding each Quality of Service dimension are described by separate utility functions. The total system quality of service utility is defined as a linear combination of the discrete form utility functions. The paper presents distributed algorithms to iteratively optimize task agents and resource agents functioning as sub-problems of the grid resource QoS scheduling optimization. Such constructed resource scheduling algorithm finds a multiple quality of service solution optimal for grid users, which fulfils some specified user preferences. The proposed pricing based distributed iterative algorithm has been evaluated by studying the effect of QoS factors on benefits of grid user utility, revenue of grid resource provider and execution success ratio.

The paper presents an evaluation of several part-of-speech taggers, representing main tagging algorithms, applied to corpus of frequency dictionary of the contemporary Polish language. We report our results considering two tagging schemes: IPI PAN positional tagset and its simplified version. Tagging accuracy is calculated for different training sets and takes into account many subcategories (accuracy on known and unknown tokens, word segments, sentences etc.) The comparison of results with other inflecting and analytic languages is done. Performance aspects (time demands) of used tagging tools are also discussed.

``Echo State'' neural networks (ESN), which are a special case of recurrent neural networks, are studied with the goal to achieve their greater predictive ability by the correction of their output signal. In this paper standard ESN was supplemented by a new correcting neural network which has served as an autoregressive predictor. The main task of this special neural network was output signal correction and therefore also a decrease of the prediction error. The goal of this paper was to compare the results achieved by this new approach with those achieved by original one-step learning algorithm. This approach was tested in laser fluctuations and air temperature prediction. Its prediction error decreased substantially in comparison to the standard approach.

A new approach for inducing decision trees is proposed based on the Variable Precision Rough Set Model. From the rough set theory point of view, in the process of inducing decision trees with evaluations of candidate attributes, some methods based on purity measurements, such as information entropy based methods, emphasize the effect of class distribution. The more unbalanced the class distribution is, the more favorable it is. The rough set based approaches emphasize the effect of certainty. The more certain it is, the better. The criterion for node selection in the new method is based on the measurement of the variable precision explicit regions corresponding to candidate attributes. We compared the presented approach with C4.5 on some data sets from the UCI machine learning repository, which instantiates the feasibility of the proposed method.