In Persian, the grammatical particle ezafe connects two words. Ezafe is one of the salient factors in Persian phonology and morphology to understand the meaning of a sentence completely and truly, whereas it is not usually written in sentences, resulting in mistakes in reading complex sentences and errors in natural language processing tasks. Therefore, recognizing words that need Ezafe at the end of themselves, is a major factor to improve the performance of a variety of NLP-based systems such as a Text TTSsystem. Because in Persian TTS systems without an Ezafe recognition module cannot make Ezafe constructions to read the text correctly and does not recognize the relations between the words. As Transformer-based methods shows state-of-the-art results in lots of NLP tasks, in this paper, we experiment ParsBERT in the task of ezafe recognition. The latter earning 2. 68% better F1-score than the prior state-of-the-art, we obtain the most advantageous outcomes.

Personality detection from social media text is well-established in English but remains underexplored for low-resource languages like Persian, despite its relevance for cultural policymaking, mental health monitoring and cybersecurity in the Iranian context. This study introduces an advanced transformer-based framework for detecting MBTI personality types from Persian Twitter (X) posts. To overcome data scarcity and class imbalance, we construct a large augmented dataset by combining the native dataset with translated multilingual tweets and applying synonym replacement augmentation. We evaluate over 10 transformer models, including Persian-specific variants (ParsBERT-Peymaner, DistilBigBird-fa) and multilingual baselines, using two user-level aggregation strategies: a flat concatenation approach and a novel Multi-Level Tweet Encoder (MLTE) that aggregates tweet embeddings via supervised contrastive learning. Robustness is further improved through a Hybrid Confidence-Selective Ensemble with dynamic model selection, calibrated weighting, and adaptive sample-wise fusion. Results show that Persian-native models consistently outperform multilingual ones. The MLTE framework achieves an average accuracy of 0. 6891, ROC-AUC of 0. 7220, and Macro F1-score of 0. 6069 on the combined dataset, surpassing flat ensembles and classical ML baselines. N/S dimension is the easiest to predict, while J/P remains the most challenging. Explainability via SHAP confirms the models rely on psychologically interpretable linguistic cues. This work contributes a large combined Persian MBTI dataset, the MLTE aggregation framework and a dynamic ensemble strategy, addressing key gaps in low-resource Persian NLP.

The rapid development of generative audio technology has brought unprecedented security and media authenticity challenges with human-like realistic deepfake speech. Recent transformer-based deepfake audio detection systems improvements are exhaustively investigated here, from traditional transformer models to self-supervised models such as WavLM, XLS-R, and Whisper. A review of fifteen recent papers is performed to locate the movement away from traditional, hand-engineered features to end-to-end, data-driven systems as the current state-of-the-art in speech, singing voice, and virtual reality contexts. Special attention is paid to publicly available datasets, large-scale data augmentation, and multimodal fusion to cope with the evolving threat environment. Comparative results are provided based on Equal Error Rate (EER) to document the technical advancements as well as the still-existing challenges towards generalizability to unseen attacks and different real-world cases. Integrating the most recent progress and open research problems, including those related to interpretability, adversarial robustness, and deployment efficiency, this survey portrays the future direction and existing challenges in deepfake audio detection research in a clear manner.

The proliferation of social networks and digital messaging platforms has significantly increased the spread of rumors and fake information within organizational and media environments. The Persian language, characterized by its orthographic diversity, rich morphological structure, and extensive use of colloquial expressions, presents unique challenges that necessitate a native framework for automated fake content detection. This study proposes a framework based on natural language processing and machine learning to classify Persian texts into three distinct categories: genuine content, fabricated information, and rumors. A dataset comprising 12, 400 Persian messages was collected and subjected to linguistic preprocessing. Three models-SVM, LSTM, and ParsBERT-were subsequently evaluated using 5-fold cross-validation. The results demonstrate that the ParsBERT model significantly outperforms classical models, achieving an F1-score of 0. 91 (p < 0. 05). These findings highlight the potential of transformer-based models for integration into early warning systems for organizational information security.