UNIVERSITY OF TAMPERE DEPARTMENT OF COMPUTER SCIENCE P.O. Box 607 SF-33101 Tampere, Finland Listed below are some of the latest technical reports of our department. Single PAPER copies are available free of charge. Quantities are limited and orders will be filled as long as reports are available. E-mail orders: em@cs.uta.fi. Please, do not forget to give your postal address!! A-1992-7 Kai Korpimies, A connectionist tutoring system. November 1992. Abstract. In this report, a connectionist tutoring system for English verb conjugation is presented. In the system both the expert and the student are modelled with connectionist networks. The architecture of the networks is borrowed from a study by Rumelhart and McClelland (1986), whereas the theoretical framework is largely provided by Smolensky's PTC theory (1988). The focal point of the report is in the comparison of the knowledge which is implicitly represented in the networks. A method is proposed, based on studying how the networks process individual examples. A-1992-5 Kai Koskimies and Juha Vihavainen, Incremental parser construction with metaobjects. November 1992. Abstract. The construction of an object-oriented recursive descent parser is studied. A program is modelled by representing each nonterminal symbol as a class. To support interactive, incremental parser construction, it is required that a modification in the definition of a nonterminal has minimal effects on the classes of other nonterminal symbols. An object- oriented parsing method based on metaobjects and lazy recursive descent technique is developed. It is shown that Eiffel allows a pseudo-incremental solution in which the changes propagate only to the next superclass level, while C++ allows fully incremental solution. A-1992-4 Pertti Jarvinen, On research into individual and computing systems. September 1992. Abstract. In order to improve quality of information systems research we must try to select an adequate research approach. When an information system consists of hardware, software and users, we have to consider every component as research objects. Their behaviour is then important. Hardware and software normally behave deterministicly. We can therefore predict their behaviour. But users do not always behave deterministicly. They have their own will and we cannot predict their behaviour. This may recommend different research approaches for computing systems on one hand and for individuals for the other hand. This fact will be demonstrated by taking two studies: a controlled experiment and a survey, and by considering them from different points of view: a) view of human being, b) horizon, c) dynamic system and d) paradigm. In two studies evaluated here the deterministic view were applied, although the voluntaristic view is considered to be more adequate. The causal models (horizon) were applied, although teleological explanations, hermeneutics and phenomenology seem to be more adequate. The human beings were also considered to behave as nilpotent systems, although the theory of dynamic systems supports such a view that they should be considered as self-steering systems. The meaning paradigm should be preferred instead of the behavioristic paradigm applied in those two studies. A-1992-3 Kalervo Jarvelin and Timo Niemi, General value conversion and aggregation operations: definition and integration with relational, entity-based and deductive data retrieval techniques. July 1992. Abstract. Existing DBMS's do not support sufficiently advanced information retrieval in heterogeneous fact database environments. From the user viewpoint this means that many information needs cannot be satisfied solely by traditional fact database operations. Transitive computation, multi-level aggregation and value conversion are frequently needed together with traditional operations. The possibilities for providing these capabilities in fact database management systems based on the relational data model are considered in this paper. It is important that the extensions for advanced data retrieval are made in a uniform way with other relational processing. This means that, on one hand, new relational operations are developed, and on the other hand, non-relational operations are integrated with relational processing via predicates expressed within relational operations. In this paper, relational algebra is extended by two generalized relational operations: one for multi-level aggregation and the other for value conversion. A set of non-relational operations (called deductive operations) for performing transitive computation is also introduced. User's query formulation can also be facilitated by providing him with an entity-based data retrieval operation on a high abstraction level. Such a high-level entity-based data retrieval operation is also introduced. The value conversion operation provides unit of measurement-related transparency. It supports very versatile conversion (including conversion of compound attributes) and checks automatically the derivability of conversion requests. The conversion expressions require minimal information from the user. The aggregation operation provides aggregation level transparency. It supports, among others, multiple layered aggregation levels and hierarchical reclassification of the classification attributes determining the aggregation levels. In the data aggregation operation, the functional dependencies between the source and result relations are connected in a complex way. Both operations are defined in this paper in an exact way so that they construct both the instances and the schemas, including functional dependencies, of the result relations. Complex data retrieval requires that value conversion, aggreerations and entity-based data retrieval operation are integrated with traditiontional operations. In this paper, a query language for advanced information retrieval consisting of these operations is developed. This language allows the intermixing of these operations with each other without limiting the nesting levels. Special attention is paid to the structures, primitives and principles in terms of which the operations and the query language can be implemented. All these aspects are defined formally, in a functional way. In other words, the definition is independent of any programming language. In addition, the concretization of these aspects in a prototype system based on Prolog-language and a workstation environment are considered. A-1992-2 Pertti Jarvinen, On purchasing process of a software package - how to teach it? April 1992. Abstract. In the software industry there is a mass production of packages for some well-defined purposes, as for text processing, network management, book-keeping, budgeting etc. Different suppliers produce packages intended for the similar tasks, and they have both the same and slightly differing features. The proportional importance of software costs compared with hardware costs is still increasing. Hence, selection of the package is a problem proper where particular knowledge and skills are needed. The course on "purchasing of a software package" was organized by this author in two last springs in the University of Tampere. The course was intended to students in computer science, but one third of the participants were from other disciplines, called application sciences. This fact gave an opportunity to form groups with three students, two from computer science and one from application sciences. The latter person played an important role, when a group defined its needs and requirements. Each group had chance to define its purchasing order, i.e. a software package to be purchased. Some of the orders were real ones and some others tentative ones. The local software companies allowed the groups to visit and to negotiate whether the package offered was suitable to the purposes defined by the group. After collecting the data about candidate packages the student groups made a comparison and then recommended one of the packages to be purchased. The groups also considered the ready made contracts and planned their proposals for changes and amendments to the conventional contracts. In this paper I shall describe the pros and cons of the approach outlined above. I will also give short descriptions of the reports prepared by the groups, actions to be taken and evaluation of various features of the course. The students expressed their views on teaching and working modes, on assessment of learning and on the whole approch. Students' evaluation was rather positive. This strongly supports the approach selected. The recommendations for improving our approach are minor and concern such wishes as changing starting time (8 o'clock) and usage of teaching material. A-1992-1 Pertti Jarvinen, Impacts of electronic markets on work. January 1992. Abstract. The domain of electronic markets are increasing in relation to electronic hierarchies. Hence it is important to analyze possible impacts of electronic markets on a human being at work. A man can be in different positions concerning electronic markets: 1. a developer of an inter-organizational information system (IOS), 2. a user (a seller or a buyer) of electronic markets, and 3. an object of markets, i.e. a specialist whose services are sold on electronic markets. First, to construct the IOS a developer has to perform particular tasks: requirements analysis, design, implementation and thereafter often also maintenance. Second, a user of electronic markets can do different things depending on whether she/he is working in a producer company, in a distributor firm, in a buyer company, in a network provider company or in financial services. Third, services offered by a human being on electronic markets should inform potential buyers about her or his service. On the other hand a buyer must describe which kind of service she or he needs.firm, in a buyer company, in a network provider company or in financial services potential buyers about her or his service. On the other hand a buyer must describe which kind of service she or he needs. In the first position, the idea of the IOS means that the production/ consumption chain is lengthened. The developer of the IOS must master this chain over two or more organizations. Hence, she or he must be more competent than nowadays. Concerning the second position, a user of electronic markets, by definition, will lose social contacts. Her variety to communicate will diminish, because she or he must use the standardized product/service description language and/or the information retrieval language at hand. In the third position, the services offered by a certain expert must be described as in detail as possible. But this creates a dilemma, if a task can be described in elementary operations, it could be automated, i.e. the task is no more necessarily performed by a human being but by a computer or by an automaton. A-1991-4 Esa Helttula, Manipulating disjoint sets: algorithms, testing, and animation, June 1991. Abstract. An analysis and an animation program of algorithms for the union- find problem is presented. The analysis program can be used to graphically compare the running times of different algorithms by counting the number of basic operations and assigning costs for them. It can also be used to study how the data structures of those algorithms change during execution. All classic algorithms and some new variations are tested empirically. The variations and some not optimal algorithms are found to outperform the classic algorithms in some cases. The animation program gives the user the possibility to try out different classic union-find algorithms and union- find-deunion algorithms. A-1991-3 Pertti Jarvinen, On approaches in information systems research, June 1991. Abstract. The domain of information systems (IS) research is very wide. The components of an IS and the main stages in the IS life cycle are different by nature. There are deterministically behaving hardware and software components. The same cannot be said about users, i.e. about conscious human beings having free will. The use and the maintenance of the information system are not the only stages or processes concerned, but we are also interested in how to design and implement a new information system. This means that in addition to discovering a current reality we are also studying how to construct a new reality. The one and same research approch is not valid for doing research work concerning different components and processes. In the recent research of central topics as: causality, determinism, dynamical systems and theories of acts, some fundamental similarities and differences are found. In various taxonomies of IS research there are different classifications of the legitimated approaches. A reader may have difficulties to understand whether the difference between two particular approaches is great or slight, and whether a certain approach is applicable to a particular purpose. Based on those fundamental similarities and differences of the central topics our aim is to show a) which research approaches are in a certain sense "relatives" and which are not, and b) which research approach is suitable for a particular purpose. Approaches are broad-mindedly classified to four "families": Firstly, causal models, statistics, field test and laboratory experiment seem to belong to the category of theorem-proving causal approaches; secondly, case study, action research, hermeneutics and phenomenology respectively to the category of sensitive exploratory approaches catching a current reality. Thirdly, we identify mathematical approaches containing e.g. formal methods and cybernetics. Fourthly, there are also constructive approaches suitable for studying how to construct a new reality. Hopefully these families or groupings of research approaches might increase the awareness of the applicability of a particular approach. This is very important in the IS society emphasizing methodological pluralism. (Presented at the IRIS'14 in Umea (Sweden), August 12-14,1991.) A-1988-10 Hannu Kangassalo, COMIC - A system for conceptual modelling and information construction, December 1988.