Publication Search

Document-level sentiment analysis assigns a single polarity label to an entire review, often obscuring opinion diversity within multi-sentence submissions. This limitation is particularly evident in reviews of multi-service platforms, where users frequently express heterogeneous opinions toward different aspects of the platform in the same review. To address this challenge, this study proposes a sentence-level sentiment analysis framework for Indonesian Gojek app reviews collected from the Google Play Store. The proposed framework introduces a two-stage segmentation strategy that combines punctuation-aware rules with conjunction-aware splitting based on coordinating and adversative conjunctions (e.g., tapi [but], padahal [even though]) to identify opinion boundaries and decompose mixed-sentiment reviews into independently classifiable sentence units. A total of 14,730 raw reviews collected between May and July 2025 were subjected to data cleaning and quality filtering, resulting in 7,187 valid reviews that were further segmented into 14,187 sentence-level instances. Each instance was manually annotated by three annotators using a four-class labeling scheme consisting of app-positive, app-negative, app-neutral, and service categories. Sentiment-level inter-annotator agreement, computed on the subset of instances unanimously categorized as app-related by all three annotators (n = 4,384), achieved substantial agreement (Fleiss'  = 0.636). Hyperparameter optimization was conducted using Optuna with the Tree-structured Parzen Estimator (TPE) sampler across four experimental scenarios. The best performance was achieved by IndoBERTweet under Stratified K-Fold evaluation, attaining an accuracy of 0.751 and a macro F1-score of 0.729, outperforming all IndoBERT configurations. The results demonstrate the effectiveness of domain-adaptive pre-training on informal Indonesian text and highlight the value of conjunction-aware segmentation for preserving fine-grained opinion structures in mixed-sentiment reviews. These findings suggest that domain-aligned language representations provide a practical and effective solution for sentence-level sentiment analysis of Indonesian app reviews.

 Cryptocurrency price prediction, particularly for highly volatile assets like Solana (SOL), is a crucial challenge in time series data analysis in digital finance. This study aims to compare the performance of the XGBoost machine learning algorithm with the Temporal Fusion Transformer (TFT) deep learning model in predicting Solana's daily closing price. The dataset used consists of historical Solana price data and network fundamentals data in the form of Total Value Locked (TVL). The research process includes data preprocessing, dividing training and test data, model training, and evaluation using the Root Mean Squared Error (RMSE) metric. The results show that using the same-day price feature has the potential to cause target leakage, resulting in invalid prediction accuracy. In testing using pure historical data without data leakage, the XGBoost model performed better than TFT with an RMSE of 4.27, while TFT produced an RMSE of 18.59. Furthermore, the integration of network fundamentals data in the form of TVL did not improve prediction accuracy and even caused a decrease in performance for the XGBoost model with an RMSE of 7.10. The results of this study show that the use of historical price action features is more effective than fundamental network indicators for short-term daily Solana price predictions.

Sentiment analysis for Indonesian regional languages faces two persistent challenges: labeled training data is extremely limited for most regional varieties, and transformer models pre-trained on Bahasa Indonesia do not generalize reliably to languages with substantially different morphological structures. Prior work on the NusaX benchmark has primarily relied on direct fine-tuning, treating each regional language independently and without exploiting linguistic proximity between related languages as a transfer signal. This paper proposes Language-Similarity-Guided Transfer (LSGT), a sequential fine-tuning strategy that first adapts a pre-trained model to a pivot language selected using character trigram similarity, followed by fine-tuning on the target language. Four transformer models are evaluated across all 12 NusaX languages using the official train/validation/test splits: IndoBERT, NusaBERT, mBERT, and XLM-R. Performance is evaluated using four metrics: accuracy, macro F1, macro precision, and macro recall. Experimental results show that LSGT improves macro F1 in 44 of 48 model-language combinations, demonstrating that the fine-tuning strategy itself is a major factor in low-resource cross-lingual sentiment classification. XLM-R benefits most strongly from LSGT, achieving an average improvement of +0.137 macro F1 and a peak gain of +0.298 on Madurese. SHAP-based token attribution analysis further reveals that predictions rely heavily on named entities and domain-specific nouns rather than sentiment-bearing vocabulary, indicating a dataset-level bias inherited from the original SmSA corpus and propagated through the NusaX translation pipeline.

The increasing sophistication of malware has rendered traditional signature-based detection methods insufficient, necessitating behavior-driven and adaptive analytical frameworks. This study presents a sequential deep learning framework that models system-level API call sequences as structured linguistic representations for behavioral malware detection. Unlike conventional comparative studies, this work systematically evaluates recurrent and attention-based architectures under controlled experimental conditions, with a particular focus on generalization performance and overfitting mitigation. Two neural architectures, a Long Short-Term Memory (LSTM) network and a Transformer-based attention model, are trained on publicly available API call sequence data for binary classification of malicious and benign executables. Beyond standard accuracy metrics, the study further examines model stability, convergence behavior, and the impact of long-range dependency modeling on detection robustness. Experimental results demonstrate that the Transformer architecture achieves superior performance, attaining 95.54% classification accuracy and consistent improvements in precision, recall, and F1-score, indicating a stronger ability to capture complex behavioral dependencies. These findings highlight the effectiveness of attention mechanisms in behavioral malware modeling and provide empirical evidence that NLP-inspired architectures offer a robust and scalable approach for real-world cybersecurity applications.

Three-phase induction motors are essential components in industrial cooling systems that require reliable overcurrent protection to maintain operational continuity and prevent equipment damage. In the Cooling Tower Pump panel at PT DongJin Indonesia, a limitation was identified where the terminal of the Electronic Overcurrent Relay (EOCR) could not accommodate the 50 mm² power cable, preventing direct current measurement. This study aims to analyze the effectiveness of integrating a 200:5 A Current Transformer (CT) with the EOCR as an adaptive protection solution that is both safe and efficient. The research employs a case study approach through field observation, motor current measurement using a Fluke 303 Clamp Meter, and descriptive analysis of the recorded data. The results indicate that the three motors operate under stable load conditions, with current values ranging from 236.7 A to 275.7 A, while the secondary current detected by the EOCR ranges from 3.7 A to 3.9 A, consistent with the CT transformation ratio. During the starting phase, the current surged to 600.4 A without causing false tripping, demonstrating that the EOCR effectively distinguishes temporary inrush current from actual fault conditions. Therefore, the integration of CT–EOCR is proven to enhance measurement safety, maintain motor operational stability, and support efficient system maintenance in industrial environments.

Inconsistent terminology across cybersecurity frameworks undermines global governance and interoperability. The National Institute of Standards and Technology Cybersecurity Framework (NIST CSF 2.0) and ISO/IEC 27001:2022 share similar objectives but diverge semantically in defining risk, control, and resilience. This semantic gap causes difficulties in compliance mapping and automated policy translation. Research Objectives: This study aims to analyze the semantic similarity and divergence between NIST and ISO/IEC 27000 terminologies, identify conceptual structures influencing interoperability, and propose an AI-assisted foundation for harmonizing cybersecurity language globally. Methodology: A mixed-method semantic comparative design integrates Natural Language Processing (NLP) and ontology mapping. Using the nist_glossary.csv dataset and ISO vocabularies, terms were normalized and analyzed via cosine similarity using sentence-transformer embeddings. Ontological alignment was visualized through the Semantic Threat Graph (STG) and validated by certified experts using Cohen’s Kappa reliability tests. Results: From 672 term pairs, results show 40.9% high semantic equivalence, 38.8% partial overlap, and 20.3% semantic divergence. Strongest alignment appears in “Protect” and “Identify” domains, while divergences occur in governance and recovery-related terms. Ontology mapping revealed three conceptual clusters—Risk Governance, Technical Safeguards, and Organizational Readiness. Conclusions: Findings confirm a 79.7% total semantic alignment, indicating strong potential for harmonizing global cybersecurity standards. The study contributes an empirical model combining computational linguistics and AI-based ontology mapping to establish semantic interoperability, enabling unified cybersecurity governance and AI-driven compliance automation. Keywords: Semantic Interoperability; Ontology Mapping; Cybersecurity Frameworks; Terminology Alignment; AI Harmonization

Prompt injection has emerged as a critical security threat for Large Language Models (LLMs), exploiting their inability to separate instructions from data within application contexts reliably. This paper provides a structured review of current attack vectors, including direct and indirect prompt injection, and highlights the limitations of existing defenses, with particular attention to the fragility of Known-Answer Detection (KAD) against adaptive attacks such as DataFlip. To address these gaps, we propose a novel, hybrid, multi-layered detection framework that operates in real-time. The architecture integrates heuristic pre-filtering for rapid elimination of obvious threats, semantic analysis using fine-tuned transformer embeddings for detecting obfuscated prompts, and behavioral pattern recognition to capture subtle manipulations that evade earlier layers. Our hybrid model achieved an accuracy of 0.974, precision of 1.000, recall of 0.950, and an F1 score of 0.974, indicating strong and balanced detection performance. Unlike prior siloed defenses, the framework proposes coverage across input, semantic, and behavioral dimensions. This layered approach offers a resilient and practical defense, advancing the state of security for LLM-integrated applications.

PLN (Persero) is the state-owned company responsible for supplying electrical energy in Indonesia and plays a crucial role in ensuring the reliability and continuity of power distribution. Therefore, all electrical equipment, particularly power transformers, must be maintained in safe and reliable operating conditions. Transformers function to change voltage levels from high to low or vice versa and are essential components in the electric power distribution system. One important transformer component that requires periodic maintenance is the On Load Tap Changer (OLTC). To maintain transformer performance and extend its service life, routine maintenance activities such as OLTC oil replacement are necessary. PT. PLN (Persero) ULTG Rangkasbitung conducts regular OLTC oil replacement twice a year as part of its preventive maintenance program. Transformer oil serves as both an insulating medium and a cooling agent; therefore, its condition greatly affects transformer reliability and operational safety. Degraded oil quality can cause insulation failure and reduce transformer efficiency. This study aims to describe the procedure and implementation of On Load Tap Changer oil replacement at the transformer substation of PT. PLN (Persero) ULTG Rangkasbitung. The research method used is field research, carried out through direct observation and interviews with maintenance personnel. The results of this study are expected to provide a clear understanding of the OLTC oil replacement process, support proper maintenance practices, and emphasize the importance of transformer maintenance in ensuring the reliability and sustainability of the electric power system.

Distribution transformers play a crucial role in delivering electrical energy from the distribution system to consumers to ensure power quality and supply continuity. However, in practice, overload conditions often occur due to increasing demand and load growth that exceed the transformer’s rated capacity. This situation can lead to reduced efficiency, increased power losses, and accelerated equipment aging. This study aims to analyze the performance of the CMY distribution transformer at PT PLN (Persero) ULP Labuan, which operates beyond its nominal capacity, and to propose an alternative solution through transformer mutation, namely the replacement of the existing unit with a transformer of more appropriate capacity based on load analysis results. The Least Square Method is employed to predict future load growth and determine the projected time when the transformer will again experience overload after the mutation. The results indicate that the existing 100 kVA transformer is overloaded and should be replaced with a 160 kVA unit. After the mutation, the loading percentage decreases significantly, the transformer’s lifespan is extended, and the reliability of the distribution system improves. Furthermore, the Least Square prediction suggests that the new transformer may experience overload again in future years if no further planning is carried out. Therefore, transformer mutation can be considered an effective and medium-term solution to enhance and maintain the reliability of the electrical distribution system within the operational area of PT PLN (Persero) ULP Labuan.

Distribution transformers play a critical role in delivering electrical energy from medium-voltage networks to low-voltage consumers. At ULP Medan Kota, several distribution transformers have been operating with loads exceeding 80% of their nominal capacity, posing risks of overloading, efficiency reduction, and equipment failure. This study aims to analyze the performance of distribution transformers based on actual load data and evaluate mitigation strategies through the implementation of additional parallel transformers (trafo sisip). The methodology includes data collection, load and current calculation, and simulation of load distribution after transformer insertion. The results show that the installation of trafo sisip reduces the load on the main transformer by approximately 50% and significantly lowers the current to safer levels. Moreover, placing the trafo sisip at an optimal position minimizes voltage drop to as low as 0.0745 Volts. Therefore, the addition of trafo sisip is proven to enhance the reliability, efficiency, and operational life of the power distribution system at ULP Medan Kota.

The increasing volume of online hotel reviews provides valuable insights into customer perceptions but poses challenges for manual analysis due to its unstructured nature. This study aims to compare the performance of Recurrent Neural Network (RNN) and Bidirectional Encoder Representations from Transformers (BERT) in hotel review sentiment analysis. A total of 20,491 TripAdvisor hotel reviews were classified into three sentiment categories: negative, neutral, and positive. The research methodology includes text preprocessing, stratified data splitting, class imbalance handling using Random Over-Sampling, tokenization, and supervised model training. Model performance was evaluated using a confusion matrix and classification metrics. The results indicate that BERT outperforms RNN, achieving an accuracy of 80.54%, while RNN reached 62.21%. BERT demonstrated superior capability in capturing contextual and semantic information in hotel reviews. These findings suggest that transformer-based models are more effective for sentiment analysis of complex textual data in the hospitality domain and can support data-driven service improvement strategies.    

This study conducts sentiment analysis of Indonesian user reviews of the CapCut application using IndoBERT and compares two evaluation schemes: a single 80/20 train–test split and stratified 5-fold cross-validation (k=5). A total of 1,048,575 reviews were collected from the Google Play Store through web scraping and labeled into three sentiment classes based on rating: negative (1–2), neutral (3), and positive (4–5). After preprocessing—cleaning, case folding, banned-word removal, normalization—and duplicate removal, 517,962 reviews were retained. IndoBERT Base P1 was fine-tuned using fixed hyperparameters (batch size 32, learning rate 2e-5, up to 4 epochs, early stopping patience 2), while undersampling was applied to the training set to address class imbalance. Performance was assessed using accuracy, precision, recall, F1-score, and ROC-AUC, supported by confusion matrix and ROC-curve visualizations. The single split achieved an accuracy of 0.756, whereas cross-validation produced a mean accuracy of 0.740. Across both schemes, the positive class achieved the best performance (F1-score 0.850; ROC-AUC 0.918–0.919), while the neutral class remained the most challenging (precision 0.198–0.206; F1-score 0.280–0.283). Overall, cross-validation is recommended for reporting because it reduces dependence on a single partition and provides a more representative estimate across multiple splits.

: The rapid expansion of cybersecurity standards and threat intelligence frameworks has led to significant semantic fragmentation among security terminologies, hindering effective information retrieval and interoperability across systems. Traditional keyword-based search approaches are inadequate for capturing the contextual meaning of security terms, particularly within formal frameworks such as NIST, MITRE ATT&CK, and CWE. This study addresses this challenge by proposing CyberBERT, a transformer-based semantic search framework designed to align cybersecurity terminologies through deep contextual representation and ontology-driven reasoning. Research Objectives: The primary objective of this research is to develop a semantic retrieval model capable of understanding conceptual relationships between security terms beyond lexical similarity. Methodology: The proposed methodology fine-tunes a BERT-based model on the NIST Glossary corpus using a combination of masked language modeling and triplet loss objectives to generate discriminative semantic embeddings. These embeddings are further aligned with cybersecurity ontologies, including MITRE ATT&CK and CWE, to enhance semantic consistency and explainability. Semantic retrieval is performed using cosine similarity within a 768-dimensional embedding space and evaluated using Mean Reciprocal Rank (MRR) and Precision@K metrics. Results: Experimental results demonstrate that CyberBERT achieves an MRR of 0.832, outperforming domain-adapted baselines such as SecureBERT and CyBERT. The integration of ontology alignment improves semantic accuracy by over 6%, while robustness evaluations confirm resilience against adversarial linguistic perturbations. Visualization using t-SNE reveals coherent semantic clustering aligned with the five core NIST Cybersecurity Framework functions. Conclusions: In conclusion, CyberBERT effectively bridges semantic gaps across cybersecurity terminologies by combining transformer-based contextual learning with ontological reasoning. The framework offers a robust, interpretable, and scalable solution for semantic search, supporting improved interoperability and knowledge discovery in cybersecurity operations and standards harmonization.

Abstract: The rapid diffusion of Generative Artificial Intelligence (AI) has intensified public debate regarding its benefits, risks, and societal implications. This study investigates public sentiment and thematic structures surrounding Generative AI by analyzing Twitter discourse as a representation of large-scale, real-time public perception. The research addresses two main problems: how public sentiment toward Generative AI is distributed and what dominant themes shape this perception. Accordingly, the objective is to map both emotional polarity and thematic narratives embedded in social media conversations. A computational mixed-methods approach was employed using a dataset of 12,470 tweets collected on 17 December 2024. Sentiment classification was conducted using a transformer-based DistilBERT model, while semantic representations were generated with Sentence-BERT. Topic modeling was performed using BERTopic, integrating HDBSCAN clustering and class-based TF-IDF to extract coherent and interpretable topics. Human-in-the-loop validation supported the interpretive robustness of topic labeling. The findings reveal that public sentiment toward Generative AI is predominantly positive (41.8%), particularly in relation to productivity enhancement, education, and creative applications. Neutral sentiment (31.4%) reflects informational discourse, while negative sentiment (26.8%) centers on ethical concerns, privacy risks, misinformation, and AI hallucinations. Seven dominant topics were identified, with clear topic–sentiment alignment showing optimism in utility-driven themes and skepticism in ethics- and risk-related discussions. In conclusion, public perception of Generative AI is dualistic—characterized by strong enthusiasm alongside persistent caution. These results provide empirical insights for AI governance, responsible innovation, and future research on socio-technical impacts of Generative AI. *    

Polusi udara partikulat halus (PM2,5) merupakan ancaman serius bagi kesehatan masyarakat di Kabupaten Brebes, Jawa Tengah. Faktor penyumbang utamanya adalah emisi kendaraan di jalur Pantura, aktivitas industri perikanan, serta konsentrasi tinggi selama musim kemarau (Juni–November). Tidak adanya model peramalan sub-jam yang akurat menghambat pengembangan sistem peringatan dini yang efektif. Penelitian ini mengembangkan dan mengevaluasi model deep learning berbasis Transformer untuk memprediksi konsentrasi PM2,5 dengan resolusi waktu 15 menit. Data yang digunakan berasal dari NASA GEOS-CF (band PM25_RH35_GCC) yang diakses melalui Google Earth Engine menggunakan API Python. Dataset mencakup periode 1 Januari hingga 22 November 2025, menghasilkan 7.813 observasi per jam, yang kemudian diinterpolasi linear menjadi 31.249 titik data dengan resolusi 15 menit. Arsitektur Transformer terdiri dari 3 lapis enkoder, 4 kepala perhatian multi-head, dimensi embedding 128, dimensi feed-forward 256, panjang sekuen 60 timestep, dan augmentasi fitur menggunakan rerata bergulir (*rolling mean*, jendela = 3) dan beda pertama (*first difference*). Pelatihan dilakukan dengan TensorFlow-Keras, pengoptimal Adam, penjadwal peluruhan kosinus (*cosine decay scheduler*), dan fungsi kerugian Huber. Pembagian data dilakukan secara kronologis: 70% pelatihan, 30% validasi. Evaluasi pada set uji independen (16 Agustus–21 November 2025, 9.357 observasi atau 97 hari 11 jam 15 menit) menghasilkan MAE 0,7691 µg/m³, RMSE 1,2052 µg/m³, R² 0,9945, dan *Explained Variance Score* 0,9948. Model ini mampu menggambarkan variasi diurnal dan anomali musiman secara akurat, jauh melampaui model LSTM dan GTWR konvensional. Penelitian ini memberikan kontribusi signifikan di bidang Teknologi Informasi melalui kerangka kerja pengolahan *big data* satelit untuk aplikasi lingkungan.

This study conceptually examines a self-supervised multi-scale fusion framework designed to enhance accuracy and computational efficiency in medical image segmentation, a domain where data scarcity and annotation cost remain major challenges. Traditional supervised approaches are constrained by their reliance on extensive labeled datasets, limiting applicability in real-world clinical environments. Self-supervised learning (SSL) mitigates this issue by extracting supervisory signals directly from unlabeled data, enabling the model to learn rich feature representations without human annotation. Simultaneously, multi-scale fusion architectures integrate global contextual information with fine-grained local features, supporting robust segmentation across varying anatomical structures and image resolutions. Through a qualitative methodology involving library research and content analysis, this study synthesizes state-of-the-art SSL-driven segmentation techniques and highlights how adaptive multi-scale fusion mechanisms address limitations of existing convolutional and transformer-based architectures. The analysis indicates that combining SSL and multi-scale strategies leads to more generalizable, scalable, and computationally efficient segmentation pipelines suitable for diverse medical imaging modalities. The proposed framework represents a promising direction for developing next-generation diagnostic tools capable of handling sparse labels, complex textures, and real-time deployment constraints.

Power transformer theft, a pervasive issue disrupting critical infrastructure, necessitates the development of cost-effective and energy-autonomous security solutions. This paper presents the design and implementation of a detection-focused anti-theft framework that integrates a Raspberry Pi Zero W, camera module, and passive infrared (PIR) motion sensors powered by a solar system for continuous monitoring. The system is designed for remote, off-grid deployment, utilizing a headless Raspberry Pi powered by a 5V solar panel and power bank to ensure energy autonomy. Upon motion detection, captured images are processed on the edge device using OpenCV’s Haar Cascade classifier, optimized for upper-body detection to minimize false positives and verify human presence. Captured images are processed locally on the edge device using OpenCV’s Haar Cascade classifier to confirm human presence before an alert is sent to the mobile application, emphasizing real-time operation and low latency. Once an intrusion is confirmed, the images are saved locally and uploaded via the Secure File Transfer Protocol to a custom-developed Android application. The app provides a dedicated remote monitoring interface, enabling secure file transfer and system access, while providing users with immediate notifications and image management capabilities. The system emphasizes low power consumption, real-time operation, and low deployment cost. Tests over 200 triggered events under varied environmental conditions achieved 90% detection accuracy with an average latency of 4.5 s. Solar autonomy was maintained for approximately 24 h under normal operation. It is concluded that the integration of solar power, edge computing of images, and mobile monitoring provides a feasible, scalable, and financially viable framework for securing transformers, especially in resource-constrained environments.

Power transformers are crucial in the electrical distribution system, and their operational stability is significantly affected by load imbalance among phases. Load imbalance can lead to the flow of neutral current through the neutral conductor, causing additional power losses in the transformer. This study analyzes the impact of load imbalance on neutral current and power losses at Transformer 1 (30 MVA capacity, 70/20 kV) at the Bungaran Substation. Data such as phase current, neutral current, and power losses were measured at 12:00 and 21:00. At 12:00, the transformer’s full-load current was 839.17 A with a loading of 28.44% and a load imbalance of 0.74%, resulting in a neutral current of 4.36 A (1.83% of load current). The power loss due to neutral current was 12.64 W (4.36×10-5 %), and the loss due to neutral current flowing to the ground was 760 W (2.62×10-3 %). At 21:00, the full-load current decreased to 834.46 A, with a loading of 29.36% and a higher load imbalance of 1.36%. This caused a neutral current of 7.94 A (3.24%), with a power loss of 41.90 W (1.43×10-4 %) and a ground power loss of 2.52 W (8.60×10-3 %). The power losses were minimal compared to the transformer’s capacity, having little effect on system efficiency. However, maintaining load balance is essential for system efficiency and transformer longevity.

This study presents a Systematic Literature Review on Artificial Intelligence (AI) and Natural Language Processing (NLP) applications for customer support automation and digital service optimization. The review follows the PRISMA framework to ensure methodological rigor and transparency, focusing on literature published between 2020 and 2025 from the Scopus database. The findings reveal that AI-driven technologies, including Machine Learning, Deep Learning, and Large Language Models, have significantly improved efficiency, response time, and customer satisfaction in customer support and digital service. Common NLP applications include sentiment analysis, ticket classification, and automated response generation. Among these, hybrid and transformer-based models demonstrate superior accuracy and contextual understanding compared to traditional algorithms. However, several challenges persist, including data quality limitations, privacy and security concerns, algorithmic bias, and linguistic ambiguities such as sarcasm and negation. Moreover, issues related to trust and ethical adoption continue to influence user acceptance of AI systems. This review provides a comprehensive synthesis of current methodologies, trends, and research gaps, offering insights for future studies to develop explainable, secure, and human-centered AI systems that enhance the sustainability and transparency of digital customer support services.

This community service program was carried out with the aim of enhancing teachers’ competencies in utilizing Artificial Intelligence technology, particularly the Generative Pre-trained Transformer (GPT), as a tool for developing digital teaching materials. The background of this activity stems from the limited understanding of teachers in applying AI technology to support 21 st-century learning. The implementation method involved intensive training that included an introduction to GPT concepts, hands-on practice in creating digital modules, designing evaluation questions, and simulating the use of teaching materials in the classroom. The activity was attended by several teachers from SMP Zawiyyah Darussalami, who demonstrated high enthusiasm throughout the sessions. Evaluation through pre-test and post-test results showed a significant improvement in teachers’ understanding and skills in using GPT. Thus, this community service activity successfully provided a positive impact by improving teachers’ competencies and opening opportunities for GPT utilization as an innovation in digital learning within schools.