Iskanje po repozitoriju

A+ | A- | | SLO | ENG

Povzetek: The proliferation of fake news across social media, headlines, and news articles poses major challenges for automated detection, particularly in multilingual and cross-media settings affected by data imbalance. We propose a fake news detection framework based on LLM-driven, feature-guided text augmentation. The method generates realistic synthetic samples across languages, media types, and text granularities while preserving mean ing and stylistic coherence. Experiments with classical and transformer-based models (Random Forest, Logistic Regression, BERT, XLM-R) across social media, headlines, and multilingual news datasets show consistent improvements in performance. For inherently balanced datasets (e.g., social media), synthetic augmentation yields negligible but stable performance changes. Across imbalanced scenarios, synthetic augmentation substantially improves minority-class recall and F1-score (e.g., fake news recall from 0.57 to 0.86), while preserving majority-class performance, leading to more balanced and reliable classifiers, whereas oversampling significantly degrades results due to overfitting on duplicated language patterns. Overall, a hybrid semantic- and style-based model proves to be the most robust strategy, outperforming oversampling and matching or exceeding baseline performance across datasets
Ključne besede: fake news detection, low-resource languages, data imbalance, synthetic data generation, prompt engineering, style-based features, semantic features
Objavljeno v DiRROS: 28.04.2026; Ogledov: 171; Prenosov: 112
Celotno besedilo (1,16 MB)
Gradivo ima več datotek! Več...

Povzetek: Ethnolinguistic vitality (ELV) studies have long explored the functioning of minority communities, emphasizing the role of language and collective identity as key unifying factors. Since 2021, an ELV study in Poland has focused on collateral language communities – autochthonous groups linked to specific territories, whose languages are closely related to the dominant national language and often misclassified as dialects. These communities are typically unrecognized, and their group distinctiveness is frequently questioned. To assess their internal cohesion, a latent variable – Collective Identity – was developed within the ELV framework. This allows for evaluating whether such communities, often viewed as regional rather than ethnic groups, exhibit shared features and a sense of collective belonging. The findings indicate that collateral language communities form distinctive entities, characterized by shared language and cultural values, thus challenging assumptions about their peripheral or non-ethnic status.
Ključne besede: ethnolinguistic vitality, collateral languages, collective identity, latent variable, Poland
Objavljeno v DiRROS: 09.02.2026; Ogledov: 553; Prenosov: 167
Celotno besedilo (1,10 MB)
Gradivo ima več datotek! Več...

Povzetek: This article adopts a comparative focus on two European borderlands, that of the Sorbs in Germany and that of the Vlachs in Serbia. It identifies three dimensions of marginalization that precede the loss of ecoculturally diverse landscapes through the expansion of mining activities: linguistic difference, border proximity, and the element of a rural culture. The article argues that these are part of a discourse that aims to legitimize the processes of “wastelanding”, that is the deliberate destruction of minority heritage through environmental degradation.
Ključne besede: Sorbian minority in Germany, Vlach minority in Serbia, ecocultural heritage, environmental degradation, endangered languages
Objavljeno v DiRROS: 19.01.2026; Ogledov: 377; Prenosov: 211
Celotno besedilo (327,25 KB)
Gradivo ima več datotek! Več...