1. |
2. TF-IDF-based classification of Uzbek educational textsKhabibulla Madatov, Sapura Sattarova, Jernej Vičič, 2025, original scientific article Abstract: This paper presents a baseline study on automatic Uzbek text classification. Uzbek is a morphologically rich and low-resource language, which makes reliable preprocessing and evaluation challenging. The approach integrates Term Frequency–Inverse Document Frequency (TF–IDF) representation with three conventional methods: linear regression (LR), k-Nearest Neighbors (k-NN), and cosine similarity (CS, implemented as a 1-NN retrieval model). The objective is to categorize school learning materials by grade level (grades 5–11) to support improved alignment between curricular texts and students’ intellectual development. A balanced dataset of Uzbek school textbooks across different subjects was constructed, preprocessed with standard NLP tools, and converted into TF–IDF vectors. Experimental results on the internal test set of 70 files show that LR achieved 92.9% accuracy (precision = 0.94, recall = 0.93, F1 = 0.93), while CS performed comparably with 91.4% accuracy (precision = 0.92, recall = 0.91, F1 = 0.92). In contrast, k-NN obtained only 28.6% accuracy, confirming its weakness in high-dimensional sparse feature spaces. External evaluation on seven Uzbek literary works further demonstrated that LR and CS yielded consistent and interpretable grade-level mappings, whereas k-NN results were unstable. Overall, the findings establish reliable baselines for Uzbek educational text classification and highlight the potential of extending beyond lexical overlap toward semantically richer models in future work.
Keywords: Uzbek language, text classification, low-resource languages, TF-IDF, cosine similarity, linear regression, k-Nearest Neighbors
Published in RUP: 17.10.2025; Views: 330; Downloads: 3
 Full text (286,87 KB)
This document has more files! More... |
3. |
4. |
5. Estimation of task-related dynamic brain connectivity via data inflation and classification model explainabilityPeter Rogelj, 2025, original scientific article Abstract: Study of brain function often involves analyzing task-related switching between intrinsic brain networks, which connect various brain regions. Functional brain connectivity analysis methods aim to estimate these networks but are limited by the statistical constraints of windowing functions, which reduce temporal resolution and hinder explainability of highly dynamic processes. In this work, we propose a novel approach to functional connectivity analysis through the explainability of EEG classification. Unlike conventional methods that condense raw data into extracted features, our approach inflates raw EEG data by decomposition into meaningful components that explain processes in the application domain. To uncover the brain connectivity that affects classification decisions, we introduce a new method of dynamic influence data inflation (DIDI), which extracts signals representing interactions between electrode regions. These inflated data are then classified using an end-to-end neural network classifier architecture designed for raw EEG signals. Saliency map estimation from trained classifiers reveals the connectivity dynamics affecting classification decisions, which can be visualized as dynamic connectivity support maps for improved interpretability. The methodology is demonstrated on two publicly available datasets: one for imagined motor movement classification and the other for emotion classification. The results highlight the dual benefits of our approach: in addition to providing interpretable insights into connectivity dynamics it increases classification accuracy.
Keywords: EEG, functional connectivity, data inflation, classification, explainability, saliency maps
Published in RUP: 04.06.2025; Views: 1880; Downloads: 17
Full text (1,74 MB)
This document has more files! More... |
6. |
7. |
8. |
9. |
10. |
