Presumptions of each data analysis are data themselves, regardless of the analysis focus (visit rate analysis, optimization of portal, personalization of portal, etc.). Results of selected analysis highly depend on the quality of analyzed data. In case of portal usage analysis, these data can be obtained by monitoring web server log file. We are able to create data matrices and web map based on these data which will serve for searching for behaviour patterns of users. Data preparation from the log file represents the most time-consuming phase of whole analysis. We realized an experiment so that we can find out to which criteria are necessary to realize this time-consuming data preparation. We aimed at specifying the inevitable steps that are required for obtaining valid data from the log file. Specially, we focused on the reconstruction of activities of the web visitor. This advanced technique of data preprocessing belongs to time consuming one. In the article we tried to assess the impact of reconstruction of activities of a web visitor on the quantity and quality of the extracted rules which represent the web users’ behaviour patterns

Understanding how software works and writing a program are currently frequent requirements when hiring employees. The complexity of learning programming often results in educational failures, student frustration and lack of motivation, because different students prefer different learning paths. Although e-learning courses have led to many improvements in the methodology and the supporting technology for more effective programming learning, misunderstanding of programming principles is one of the main reasons for students leaving school early. Universities face a challenging task: how to harmonise students’ education, focusing on advanced knowledge in the development of software applications, with students’ education in cases where writing code is a new skill. The article proposes a conceptual framework focused on the comprehensive training of future programmers using microlearning and automatic evaluation of source codes to achieve immediate feedback for students. This framework is designed to involve students in the software development of virtual learning environment software that will provide their education, thus ensuring the sustainability of the environment in line with modern development trends. The paper’s final part is devoted to verifying the contribution of the presented elements through quantitative research on the introductory parts of the framework. It turned out that although the application of interactive features did not lead to significant measurable progress during the first semester of study, it significantly improved the results of students in subsequent courses focused on advanced programming.

Due to the constantly evolving social media and different types of sources of information, we are facing different fake news and different types of misinformation. Currently, we are working on a project to identify applicable methods for identifying fake news for floating language types. We explored different approaches to detect fake news in the presented research, which are based on morphological analysis. This is one of the basic components of natural language processing. The aim of the article is to find out whether it is possible to improve the methods of dataset preparation based on morphological analysis. We collected our own and unique dataset, which consisted of articles from verified publishers and articles from news portals that are known as the publishers of fake and misleading news. Articles were in the Slovak language, which belongs to the floating types of languages. We explored different approaches in this article to the dataset preparation based on morphological analysis. The prepared datasets were the input data for creating the classifier of fake and real news. We selected decision trees for classification. The evaluation of the success of two different methods of preparation was carried out because of the success of the created classifier. We found a suitable dataset pre-processing technique by morphological group analysis. This technique could be used for improving fake news classification.

Contemporary language models heavily rely on large corpora for their training. The larger the corpus, the better a model can capture various semantic relationships. The issue at hand appears to be the limited scope of the corpora used. One potential solution to this problem is the application of data augmentation techniques to expand the existing corpus. Data augmentation encompasses several techniques for corpus augmentation. In this article, we delve deeper into the analysis of three techniques: Synonym Replacement, Back Translation, and Reduction of Function Words. Utilizing these three techniques, we prepared diverse versions of the corpus employed for training Word2Vec Skip-gram models. These techniques were validated through extrinsic evaluation, wherein Word2Vec Skip-gram models were used to generate word vectors for classifying fake news articles. Performance measures of the generated classifiers were analyzed. The study highlights significant statistically differences in classifier outcomes between augmented and original corpora. Specifically, Back Translation significantly enhances accuracy, notably with Support Vector and Bernoulli Naive Bayes models. Conversely, Reduction of Function Words (FWD) improves Logistic Regression, while the original corpus excels in Random Forest classification. The article also includes an intrinsic evaluation involving lexical semantic relations between word pairs. The intrinsic evaluation highlights nuanced differences in semantic relations across augmented corpora. Notably, the Back Translation (BT) corpus better aligns with established lexical resources, showcasing promising improvements in understanding specific semantic relationships.

Research of the techniques for effective fake news detection has become very needed and attractive. These techniques have a background in many research disciplines, including morphological analysis. Several researchers stated that simple content-related n-grams and POS tagging had been proven insufficient for fake news classification. However, they did not realise any empirical research results, which could confirm these statements experimentally in the last decade. Considering this contradiction, the main aim of the paper is to experimentally evaluate the potential of the common use of n-grams and POS tags for the correct classification of fake and true news. The dataset of published fake or real news about the current Covid-19 pandemic was pre-processed using morphological analysis. As a result, n-grams of POS tags were prepared and further analysed. Three techniques based on POS tags were proposed and applied to different groups of n-grams in the pre-processing phase of fake news detection. The n-gram size was examined as the first. Subsequently, the most suitable depth of the decision trees for sufficient generalization was scoped. Finally, the performance measures of models based on the proposed techniques were compared with the standardised reference TF-IDF technique. The performance measures of the model like accuracy, precision, recall and f1-score are considered, together with the 10-fold cross-validation technique. Simultaneously, the question, whether the TF-IDF technique can be improved using POS tags was researched in detail. The results showed that the newly proposed techniques are comparable with the traditional TF-IDF technique. At the same time, it can be stated that the morphological analysis can improve the baseline TF-IDF technique. As a result, the performance measures of the model, precision for fake news and recall for real news, were statistically significantly improved.

Currently, we are once again experiencing a frenzy related to artificial intelligence. Generative Pre-trained Transformers (GPT) models are highly effective at various natural language processing tasks. Different varieties of GPT models are widely used these days to improve productivity. Graphic departments generate art designs, developers engineer intricate software solutions, leveraging services predicated on the GPT framework, and many other industries are also following the lead and implementing these new sets of tools in their workflow. However, there are areas in natural language processing where a simple solution is often more suitable and effective than current Large Language Models. In this article, we decided to analyze and compare the practical use of one of the more popular GPT solutions, J-Large, and the simple rule-based model we implemented. We integrated these two models into the internal information system of a private company focused on communication with customers in the gaming industry. Both models were trained on the same dataset provided as a log of conversational interactions for the last two years in the given system. We observed that GPT models exhibited superior performance in terms of comprehensibility and adequacy. The rule-based models showed noticeable proficiency in handling domain-specific tasks, mainly when fed with datasets extracted from the historical communication between users and a specialized domain system, such as a customer care department. As a result, with a sufficiently tailored and specific dataset at their disposal, rule-based models can effectively outpace GPT models in performing domain-specific tasks.