Czeslaw Danilowicz

Consensus-based partitions in the space of ordered partitions

Abstract The paper presents a method of determining a representation of ordered partitions,i.e. a partition to which the sum of distance from fixed partitions is minimal. Two metrics are proposed as measures of distance between ordered partitions. The first metric is equal to the minimum number of element moves necessary to transform a partition into another one, and the second one equals the sum of weights of these element moves. Criteria of selecting best reprsentations of chosen ordered partitions are also presented.

Modelling of user preferences and needs in Boolean retrieval systems

Abstract The constant growth of the number of end-users who are deprived of the help of intermediary searchers is a source of substantial problems which have to be solved at the different levels of the communication between the human and the computer system. As far as the information retrieval system is concerned the main problem is to provide a satisfactory output to a typical query put by a typical end-user. In this article we have shown that the problems of the end-users can be efficiently solved when the additional information about user needs and preferences is introduced by means of the socalled “profiles.” The complete description of the information retrieval model, its characteristics, and algorithms for computing the profiles have been presented. Some ways of modification and forming a general model are shown at the end of the article.

Document ranking based upon Markov chains

Abstract One of the most important problems in information retrieval is determining the order of documents in the answer returned to the user. Many methods and algorithms for document ordering have been proposed. The method introduced in this paper differs from them especially in that it uses a probabilistic model of a document set. In this model documents are regarded as states of a Markov chain, where transition probabilities are directly proportional to similarities between documents. Steady-state probabilities reflect similarities of particular documents to the whole answer set. If documents are ordered according to these probabilities, at the top of a list there will be documents that are the best representatives of the set, and at the bottom those which are the worst representatives. The method was tested against databases INSPEC and Networked Computer Science Technical Reference Library (NCSTRL). Test results are positive. Values of the Kendall rank correlation coefficient indicate high similarity between rankings generated by the proposed method and rankings produced by experts. Results are comparable with rankings generated by the vector model using standard weighting schema tf·idf.

Consensus-based Methods for Restoring Consistency of Replicated Data

In this paper a consensus model for restoring consistency of replicated data is presented. It is assumed that after some time of functioning of a distributed system, versions of replicated data stored in different servers may differ from each other, and the only basis for recreating the proper data version is the set of these versions. The authors propose to determine the consensus of data versions and take it as the proper version. In this work the consensus structure as a metric space, consensus choice and its analysis, and algorithms for determining most often used consensus functions are presented.

Using User Profiles in Intelligent Information Retrieval

Personalization has been recently one of the most important features of intelligent information retrieval. An intelligent system should store information about user interests and utilize this information to deliver to the user documents he really needs. In such a system the information needs of a user should be represented by means of so called a user profile. User profiles, in other hand, should be used together with queries to sort retrieved information in such order that is adequate to user preferences. In this paper a vector-based information system model is presented, in which the user information needs and preferences (profiles) are defined and the methods for updating user profiles and automatic learning about user preferences are worked out.

A model conception for optimal scenario determination in an intelligent learning system

In this paper a conception of the model for learning scenario determination is presented. We define the learning scenario as a sequence of the hypermedia pages, representing particular knowledge units, and tests related to them. The scenario determination is a dynamic process that begins when a new student takes up a course. The opening scenario for this student is chosen as the consensus of the final scenarios of the students, who have already finished this course, and who belong to a class of the learners similar to the new one. We have elaborated the consensus‐based procedure for the scenario determination. Since this procedure operates on a set of similar learners, we have developed the conceptions of learner’s profile and students’ classification. The learner’s profile is proposed to include the attributes describing students’ personal data (as name, birthday etc.), their cognitive and learning styles as well as their usage data (represented by the learning scenarios). The students’ classification is based on a set of the basic attributes that seem to influence the learning effects. Their significance is verified during the learning process. We have also elaborated the procedure of reducing undistinguishable values of the attribute and removing useless attributes from the set of basic attributes. A learning procedure proposed, describes generally the situations when the scenario is modified, and the methods used for its modification.

Consensus Methods for Solving Inconsistency of Replicated Data in Distributed Systems

Replication of data is a popular and convenient form of data organization in distributed systems. Together with its advantages, data replication brings specific problems, which have to be solved by system designers. This paper deals with methods for resolving inconsistencies in data replication. The problem investigated in this work is: How to restore the data consistency if after some time of functioning their versions differ from each other on some sites of the system. We propose a solution of this problem by determining consensus of replicated data versions. We assume that there is a possibility to define a distance function between versions of replicated data, next different consensus choice functions are defined and analyzed. A numerical and practical example of applying these methods is also presented.

Dynamic user profiles based on boolean formulas

One of the difficulties of using Artificial Neural Networks (ANNs) to estimate atmospheric temperature is the large number of potential input variables available. In this study, four different feature extraction methods were used to reduce the input vector to train four networks to estimate temperature at different atmospheric levels. The four techniques used were: genetic algorithms (GA), coefficient of determination (CoD), mutual information (MI) and simple neural analysis (SNA). The results demonstrate that of the four methods used for this data set, mutual information and simple neural analysis can generate networks that have a smaller input parameter set, while still maintaining a high degree of accuracy.

Selection of scientific journals based on the data obtained from an information service system

Abstract A great and constantly growing number of scientific journals editions demands strict selection of journals during the planning of their purchase. The method of journal selection on the basis of the information service system data is described in this article. The primary value of a journal has been defined as an amount of the retrieved for the readers information concerning the articles published in a given journal. This parameter and the costs of subscription are the basis for journal ranking and determination of the number of copies (to be bought). This method has been verified by the use of data from the SDI system exploited at Wroclaw Technical University. Having compared the achieved results with the results of the simultaneously conducted questionnaire investigation, it has been found that there is a considerable degree of accordance between the results of journal acquisition planning based on the described method, and the demands of journal users.

Relative indexing on the basis of users' profiles

Abstract Principles for determining the profile of a system user have been presented. These principles are based on the analysis of co-occurrence of index terms in queries and pertinent documents. Moreover, a procedure for determining index term weights on the basis of user profiles has been introduced. The information value of the index term weights depends on the degree of homogeneity of the system users.