Esther Irawati Setiawan
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Esther Irawati Setiawan has a total of 75 co-authors. Here are the top co-authors by number of shared publications:
- Joan Santoso
Institut Sains dan Teknologi Terpadu Surabaya
11 shared publications - Yosi Kristian
Institut Sains dan Teknologi Terpadu Surabaya
5 shared publications Joan Santoso
Affiliation Information Technology Department, Institut Sains dan Teknologi Terpadu Surabaya, Surabaya, Jawa Timur, Indonesia
4 shared publications- Gunawan Gunawan
Institut Sains dan Teknologi Terpadu Surabaya
3 shared publications - Mauridhi Hery Purnomo
Department of Electrical Engineering, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
3 shared publications Syaiful Imron
PT Mulya Adhi Paramita, Indonesia
3 shared publications
Latest Publications
- Open Access
- 30 March 2026
- Published by Society of Visual Informatics in JOIV : International Journal on Informatics Visualization
- Vol. 10 (2) , 471
Procedural Horror Game Using Rule-Based and Markov Chain PCG with Cultural Fear Modeling
This research proposes a Procedural Content Generation (PCG) system using Markov and Rule-based models to control the rhythm of fear in 2D platformer horror games. The system integrates elements of the fear model, consisting of darkness, apparitions, sounds, and text prompts, to produce dynamic levels that maintain a cohesive gameplay flow. Unlike the PCG approach, which relies solely on probabilistic sequences, this study incorporates affective constraints directly into the Markov transition matrix to generate platform patterns that ensure structured transitions between fear-inducing and recovery areas. Implemented using the Unity game engine that produces a 2D platformer based on prefab tiles with the integration of fear model elements using cultural representations through Indonesian ghost characters, such as Kuntilanak, Pocong, and Genderuwo. The probabilistic placement of prefab tiles on 2D platform areas is performed using rule-based filtering to maintain game feasibility by enforcing platform height restrictions, fear-trigger areas, safe areas, and item-distribution areas. The evaluation, involving 30 respondents, yielded 3,000 platform transitions, which were compared with a random generator layout. The Markov and rule-based models eliminated sequential ghost-area transitions, thereby improving the stability of fear-recovery pacing compared with a random generator. Fear-triggering areas also appeared less frequently than recovery areas, with Kuntilanak Road appearing at 9.7%, Pocong Road at 8.8%, and Genderuwo Road at 4.0%.
- 30 March 2026
- Published by Society of Visual Informatics in JOIV : International Journal on Informatics Visualization
- Vol. 10 (2) , 471
Open AccessProcedural Horror Game Using Rule-Based and Markov Chain PCG with Cultural Fear Modeling
This research proposes a Procedural Content Generation (PCG) system using Markov and Rule-based models to control the rhythm of fear in 2D platformer horror games. The system integrates elements of the fear model, consisting of darkness, apparitions, sounds, and text prompts, to produce dynamic levels that maintain a cohesive gameplay flow. Unlike the PCG approach, which relies solely on probabilistic sequences, this study incorporates affective constraints directly into the Markov transition matrix to generate platform patterns that ensure structured transitions between fear-inducing and recovery areas. Implemented using the Unity game engine that produces a 2D platformer based on prefab tiles with the integration of fear model elements using cultural representations through Indonesian ghost characters, such as Kuntilanak, Pocong, and Genderuwo. The probabilistic placement of prefab tiles on 2D platform areas is performed using rule-based filtering to maintain game feasibility by enforcing platform height restrictions, fear-trigger areas, safe areas, and item-distribution areas. The evaluation, involving 30 respondents, yielded 3,000 platform transitions, which were compared with a random generator layout. The Markov and rule-based models eliminated sequential ghost-area transitions, thereby improving the stability of fear-recovery pacing compared with a random generator. Fear-triggering areas also appeared less frequently than recovery areas, with Kuntilanak Road appearing at 9.7%, Pocong Road at 8.8%, and Genderuwo Road at 4.0%.
- Open Access
- 16 July 2025
- Published by Universitas Malikussaleh in International Journal of Engineering, Science and Information Technology
- Vol. 5 (4) , 427-433
Evaluating User Experience of a Virtual Reality-Based Adaptive Learning Application on Chemical Compound Structures for High School Students
Recognizing the significant spatial visualization challenges that high school students face in understanding abstract chemical compound structures—a limitation often inherent in conventional teaching methods based on 2D diagrams—this research presents the comprehensive development and user experience (UX) evaluation of an innovative adaptive learning application in Virtual Reality (VR). The application, developed using the Unity 3D engine and configured via XR Plugin Management to ensure broad hardware compatibility, places students in an interactive virtual laboratory. Within it, students can directly manipulate meticulously designed 3D atomic models to build molecules, observe the formation of covalent and ionic bonds, and interact with dynamic chemical processes. Its key innovation is the integration of an intelligent adaptive learning algorithm, which utilizes a Firebase cloud database to analyze user performance metrics—such as accuracy, completion time, and recurring areas of difficulty. Based on this data, the system dynamically personalizes learning pathways by recommending remedial content or more challenging topics. Furthermore, assessment materials such as quizzes were efficiently generated using large language models (LLMs) to ensure relevance and quality. An in-depth UX evaluation was conducted with high school students using a mixed-methods approach, combining standardized questionnaires to quantitatively measure metrics like usability, engagement, and satisfaction, with qualitative feedback sessions for contextual insights. The results indicate a highly positive user experience; participants reported that the ability to directly manipulate molecules in 3D space significantly enhanced their conceptual understanding, bridging the gap between theory and visualization. The adaptive system was highly valued for its ability to adjust to individual learning paces, which was shown to boost confidence and reduce frustration. This research provides strong evidence that VR-based adaptive learning platforms are powerful pedagogical tools, capable of transforming chemistry education by making complex scientific concepts more accessible, engaging, and comprehensible.
- 16 July 2025
- Published by Universitas Malikussaleh in International Journal of Engineering, Science and Information Technology
- Vol. 5 (4) , 427-433
Open AccessEvaluating User Experience of a Virtual Reality-Based Adaptive Learning Application on Chemical Compound Structures for High School Students
Recognizing the significant spatial visualization challenges that high school students face in understanding abstract chemical compound structures—a limitation often inherent in conventional teaching methods based on 2D diagrams—this research presents the comprehensive development and user experience (UX) evaluation of an innovative adaptive learning application in Virtual Reality (VR). The application, developed using the Unity 3D engine and configured via XR Plugin Management to ensure broad hardware compatibility, places students in an interactive virtual laboratory. Within it, students can directly manipulate meticulously designed 3D atomic models to build molecules, observe the formation of covalent and ionic bonds, and interact with dynamic chemical processes. Its key innovation is the integration of an intelligent adaptive learning algorithm, which utilizes a Firebase cloud database to analyze user performance metrics—such as accuracy, completion time, and recurring areas of difficulty. Based on this data, the system dynamically personalizes learning pathways by recommending remedial content or more challenging topics. Furthermore, assessment materials such as quizzes were efficiently generated using large language models (LLMs) to ensure relevance and quality. An in-depth UX evaluation was conducted with high school students using a mixed-methods approach, combining standardized questionnaires to quantitatively measure metrics like usability, engagement, and satisfaction, with qualitative feedback sessions for contextual insights. The results indicate a highly positive user experience; participants reported that the ability to directly manipulate molecules in 3D space significantly enhanced their conceptual understanding, bridging the gap between theory and visualization. The adaptive system was highly valued for its ability to adjust to individual learning paces, which was shown to boost confidence and reduce frustration. This research provides strong evidence that VR-based adaptive learning platforms are powerful pedagogical tools, capable of transforming chemistry education by making complex scientific concepts more accessible, engaging, and comprehensible.
- Open Access
- 17 June 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (6) , e0324394
An image and text-based fake news detection with transfer learning
Fake news has emerged as a significant problem in today’s information age, threatening the reliability of information sources. Detecting fake news is crucial for maintaining trust and ensuring access to factual information. While deep learning offers solutions, most approaches focus on the text, neglecting the potential of visual information, which may contradict or misrepresent the accompanying text. This research proposes a multimodal classification approach that combines text and images to improve fake news detection, particularly in low-resource settings where labeled data is scarce. We leverage CLIP, a model that understands relationships between images and text, to extract features from both modalities. These features are concatenated and fed into a simple one-layer multi-layer perceptron (MLP) for classification. To enhance data efficiency, we apply LoRA (Low-Rank Adaptation), a parameter-efficient fine-tuning technique, to the CLIP model. We also explore the effects of integrating features from other models. The model achieves an 83% accuracy when using LoRA in classifying whether an image and its accompanying text constitute fake or factual news. These results highlight the potential of multimodal learning and efficient fine-tuning techniques for robust fake news detection, even with limited data.
- 17 June 2025
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 20 (6) , e0324394
Open AccessAn image and text-based fake news detection with transfer learning
Fake news has emerged as a significant problem in today’s information age, threatening the reliability of information sources. Detecting fake news is crucial for maintaining trust and ensuring access to factual information. While deep learning offers solutions, most approaches focus on the text, neglecting the potential of visual information, which may contradict or misrepresent the accompanying text. This research proposes a multimodal classification approach that combines text and images to improve fake news detection, particularly in low-resource settings where labeled data is scarce. We leverage CLIP, a model that understands relationships between images and text, to extract features from both modalities. These features are concatenated and fed into a simple one-layer multi-layer perceptron (MLP) for classification. To enhance data efficiency, we apply LoRA (Low-Rank Adaptation), a parameter-efficient fine-tuning technique, to the CLIP model. We also explore the effects of integrating features from other models. The model achieves an 83% accuracy when using LoRA in classifying whether an image and its accompanying text constitute fake or factual news. These results highlight the potential of multimodal learning and efficient fine-tuning techniques for robust fake news detection, even with limited data.
- Open Access
- 9 May 2025
- Published by Universitas Malikussaleh in International Journal of Engineering, Science and Information Technology
- Vol. 5 (3) , 316-326
Kuntilanak as a Runtime Entity: Technical Integration of Javanese Folklore Using Manga Matrix in a 2D Horror Game
In this work, Kuntilanak, a mythological creature from Javanese mythology, is used as a dynamic element in a 2D horror game to provide a technical framework for integrating culturally infused folklore into interactive gaming. The design process breaks down the character's appearance, attire, and personality into workable technical specifications using the Manga Matrix framework as a guide. With C# scripted behaviours like unexpected appearances, animation state changes (controlled by Unity's Animator Controller), audio triggers (laughing, crying), and interactive reactions to in-game objects like yellow Bamboo (for hiding) and scissors (for repelling), Kuntilanak was created as a sprite-based runtime entity inside the Unity game engine. The character can be dynamically instantiated thanks to this technical approach, which supports procedural horror encounters and is consistent with traditional narratives. The effectiveness of the suggested technological integration was validated by a quantitative assessment using a Likert scale (N=50), which showed 82.2% agreement on cultural authenticity and 79.5% on emotional impact. The findings support the methodology's capacity to turn folklore characters into functional game entities and offer a replicable model for serious games that consider cultural sensitivity. The findings support the methodology's capacity to turn folklore characters into functional game entities and provide a replicable model for serious games that consider cultural sensitivity, with direct implications for designing engaging educational experiences that promote cultural heritage preservation.
- 9 May 2025
- Published by Universitas Malikussaleh in International Journal of Engineering, Science and Information Technology
- Vol. 5 (3) , 316-326
Open AccessKuntilanak as a Runtime Entity: Technical Integration of Javanese Folklore Using Manga Matrix in a 2D Horror Game
In this work, Kuntilanak, a mythological creature from Javanese mythology, is used as a dynamic element in a 2D horror game to provide a technical framework for integrating culturally infused folklore into interactive gaming. The design process breaks down the character's appearance, attire, and personality into workable technical specifications using the Manga Matrix framework as a guide. With C# scripted behaviours like unexpected appearances, animation state changes (controlled by Unity's Animator Controller), audio triggers (laughing, crying), and interactive reactions to in-game objects like yellow Bamboo (for hiding) and scissors (for repelling), Kuntilanak was created as a sprite-based runtime entity inside the Unity game engine. The character can be dynamically instantiated thanks to this technical approach, which supports procedural horror encounters and is consistent with traditional narratives. The effectiveness of the suggested technological integration was validated by a quantitative assessment using a Likert scale (N=50), which showed 82.2% agreement on cultural authenticity and 79.5% on emotional impact. The findings support the methodology's capacity to turn folklore characters into functional game entities and offer a replicable model for serious games that consider cultural sensitivity. The findings support the methodology's capacity to turn folklore characters into functional game entities and provide a replicable model for serious games that consider cultural sensitivity, with direct implications for designing engaging educational experiences that promote cultural heritage preservation.
- Open Access
- 9 May 2025
- Published by Universitas Malikussaleh in International Journal of Engineering, Science and Information Technology
- Vol. 5 (3) , 268-277
Retrieval Augmented Generation-Based Chatbot for Prospective and Current University Students
Universities utilize chatbots as assistants for users, especially prospective and current students, to access information and answer questions with relevant answers. This study introduces a new approach to an open-source model-based QA system using Gemma2-2b-it by combining Retrieval Augmented Generation (RAG) and Fine-tuning (FT) techniques. Previously, some studies have focused on only one approach, but this study will combine and compare both methods separately. Raw conversation data from WhatsApp, the main university website, and university PDF documents are used. The Retrieval Augmented Generation Assessment (RAGAS) framework will be used to evaluate the performance of the RAG model. In contrast, precision, recall, and similarity are used to assess the comparative performance of RAG and fine-tuning. The results of the RAGAS show that RAG using the base model is better than RAG using a fine-tuned model, which has 0.78 faithfulness, 0.64 answer relevancy, 0.81 context precision, and 0.68 context recall, so the overall RAGAS Score is 0.72. The comparison of precision and recall of fine-tuning are higher than those of using RAG, but the similarity score is not much different. Furthermore, the potential improvement for RAG of this study can be increased by adding a reranking process in the retrieved context, and fine-tuning of the embedding model can also be added to increase the retrieval process's performance. In addition, further experiments on various datasets and the challenge of overfitting in fine-tuning must be overcome so that the model can also perform better generalization.
- 9 May 2025
- Published by Universitas Malikussaleh in International Journal of Engineering, Science and Information Technology
- Vol. 5 (3) , 268-277
Open AccessRetrieval Augmented Generation-Based Chatbot for Prospective and Current University Students
Universities utilize chatbots as assistants for users, especially prospective and current students, to access information and answer questions with relevant answers. This study introduces a new approach to an open-source model-based QA system using Gemma2-2b-it by combining Retrieval Augmented Generation (RAG) and Fine-tuning (FT) techniques. Previously, some studies have focused on only one approach, but this study will combine and compare both methods separately. Raw conversation data from WhatsApp, the main university website, and university PDF documents are used. The Retrieval Augmented Generation Assessment (RAGAS) framework will be used to evaluate the performance of the RAG model. In contrast, precision, recall, and similarity are used to assess the comparative performance of RAG and fine-tuning. The results of the RAGAS show that RAG using the base model is better than RAG using a fine-tuned model, which has 0.78 faithfulness, 0.64 answer relevancy, 0.81 context precision, and 0.68 context recall, so the overall RAGAS Score is 0.72. The comparison of precision and recall of fine-tuning are higher than those of using RAG, but the similarity score is not much different. Furthermore, the potential improvement for RAG of this study can be increased by adding a reranking process in the retrieved context, and fine-tuning of the embedding model can also be added to increase the retrieval process's performance. In addition, further experiments on various datasets and the challenge of overfitting in fine-tuning must be overcome so that the model can also perform better generalization.
- Open Access
- 1 December 2024
- Published by Bright Publisher in Journal of Applied Data Sciences
- Vol. 5 (4) , 1579-1596
Aspect-Based Sentiment Analysis of Healthcare Reviews from Indonesian Hospitals based on Weighted Average Ensemble
Aspect-Based Sentiment Analysis of Healthcare Reviews from Indonesian Hospitals based on Weighted Average Ensemble
- 1 December 2024
- Published by Bright Publisher in Journal of Applied Data Sciences
- Vol. 5 (4) , 1579-1596
Open AccessAspect-Based Sentiment Analysis of Healthcare Reviews from Indonesian Hospitals based on Weighted Average Ensemble
Aspect-Based Sentiment Analysis of Healthcare Reviews from Indonesian Hospitals based on Weighted Average Ensemble
- Open Access
- 3 August 2024
- Published by Green Publisher in Eduvest - Journal of Universal Studies
- Vol. 4 (8) , 7379-7398
Factors Affecting The Adoption Of Mobile Learning In Vocational High Schools And High Schools Using Extended UTAUT
M-Learning is a learning process that uses technology or mobile devices such as smartphones, tablets or wearable devices to support the learning process. This is still being done because there are many different theoretical models proposed. However, there is no model that can be generally accepted as an established theoretical model in the application of M-learning in vocational and high school education environments in Sidoarjo. This research is expected to make a significant contribution to the development of a better theoretical understanding of the determining factors that influence M-learning adoption using the Unified Theory of Acceptance and Use of The Technology (UTAUT). To collect data, researchers distributed questionnaires to respondents using Google Form. The data used were 444 M-learning users. Theoretical model research was carried out using Structural Equation Modeling (SEM) analysis, then SPSS and Amos as analysis support. There are seven factors that determine the results of acceptance of M-Learning adoption in this research, namely Facilitating Condition, Performance Expectancy, Effort Expectancy, Perceived Convenience, Social Influence, School Management Support. The six factors that show a positive and significant relationship are Facilitating Condition, Performance Expectancy, Effort Expectancy, Perceived Convenience, Social Influence, School Management Support. Perceived Convenience has the first strongest positive and significant value, and Performance Expectancy has the second strongest value. Each factor has a moderate influence on Intention to Use. This factor is the most influential in implementing M-Learning in vocational and high schools in the Sidoarjo area.
- 3 August 2024
- Published by Green Publisher in Eduvest - Journal of Universal Studies
- Vol. 4 (8) , 7379-7398
Open AccessFactors Affecting The Adoption Of Mobile Learning In Vocational High Schools And High Schools Using Extended UTAUT
M-Learning is a learning process that uses technology or mobile devices such as smartphones, tablets or wearable devices to support the learning process. This is still being done because there are many different theoretical models proposed. However, there is no model that can be generally accepted as an established theoretical model in the application of M-learning in vocational and high school education environments in Sidoarjo. This research is expected to make a significant contribution to the development of a better theoretical understanding of the determining factors that influence M-learning adoption using the Unified Theory of Acceptance and Use of The Technology (UTAUT). To collect data, researchers distributed questionnaires to respondents using Google Form. The data used were 444 M-learning users. Theoretical model research was carried out using Structural Equation Modeling (SEM) analysis, then SPSS and Amos as analysis support. There are seven factors that determine the results of acceptance of M-Learning adoption in this research, namely Facilitating Condition, Performance Expectancy, Effort Expectancy, Perceived Convenience, Social Influence, School Management Support. The six factors that show a positive and significant relationship are Facilitating Condition, Performance Expectancy, Effort Expectancy, Perceived Convenience, Social Influence, School Management Support. Perceived Convenience has the first strongest positive and significant value, and Performance Expectancy has the second strongest value. Each factor has a moderate influence on Intention to Use. This factor is the most influential in implementing M-Learning in vocational and high schools in the Sidoarjo area.
- Open Access
- 19 April 2024
- Published by LP2M Universitas Islam Negeri (UIN) Syarif Hidayatullah Jakarta in Applied Information System and Management (AISM)
- Vol. 7 (1) , 43-52
Cross Platform Waste Reuse, Reduce And Recycle Management Application With Prototyping Methodology
Cross Platform Waste Reuse, Reduce And Recycle Management Application With Prototyping Methodology
- 19 April 2024
- Published by LP2M Universitas Islam Negeri (UIN) Syarif Hidayatullah Jakarta in Applied Information System and Management (AISM)
- Vol. 7 (1) , 43-52
Open AccessCross Platform Waste Reuse, Reduce And Recycle Management Application With Prototyping Methodology
Cross Platform Waste Reuse, Reduce And Recycle Management Application With Prototyping Methodology
- Open Access
- 16 April 2024
- Published by State University of Malang (UM) in Knowledge Engineering and Data Science
- Vol. 7 (1) , 101-116
Timbre Style Transfer for Musical Instruments Acoustic Guitar and Piano using the Generator-Discriminator Model
Music style transfer is a technique for creating new music by combining the input song's content and the target song's style to have a sound that humans can enjoy. This research is related to timbre style transfer, a branch of music style transfer that focuses on using the generator-discriminator model. This exciting method has been used in various studies in the music style transfer domain to train a machine learning model to change the sound of instruments in a song with the sound of instruments from other songs. This work focuses on finding the best layer configuration in the generator-discriminator model for the timbre style transfer task. The dataset used for this research is the MAESTRO dataset. The metrics used in the testing phase are Contrastive Loss, Mean Squared Error, and Perceptual Evaluation of Speech Quality. Based on the results of the trials, it was concluded that the best model in this research was the model trained using column vectors from the mel-spectrogram. Some hyperparameters suitable in the training process are a learning rate 0.0005, batch size greater than or equal to 64, and dropout with a value of 0.1. The results of the ablation study show that the best layer configuration consists of 2 Bi-LSTM layers, 1 Attention layer, and 2 Dense layers.
- 16 April 2024
- Published by State University of Malang (UM) in Knowledge Engineering and Data Science
- Vol. 7 (1) , 101-116
Open AccessTimbre Style Transfer for Musical Instruments Acoustic Guitar and Piano using the Generator-Discriminator Model
Music style transfer is a technique for creating new music by combining the input song's content and the target song's style to have a sound that humans can enjoy. This research is related to timbre style transfer, a branch of music style transfer that focuses on using the generator-discriminator model. This exciting method has been used in various studies in the music style transfer domain to train a machine learning model to change the sound of instruments in a song with the sound of instruments from other songs. This work focuses on finding the best layer configuration in the generator-discriminator model for the timbre style transfer task. The dataset used for this research is the MAESTRO dataset. The metrics used in the testing phase are Contrastive Loss, Mean Squared Error, and Perceptual Evaluation of Speech Quality. Based on the results of the trials, it was concluded that the best model in this research was the model trained using column vectors from the mel-spectrogram. Some hyperparameters suitable in the training process are a learning rate 0.0005, batch size greater than or equal to 64, and dropout with a value of 0.1. The results of the ablation study show that the best layer configuration consists of 2 Bi-LSTM layers, 1 Attention layer, and 2 Dense layers.
- Open Access
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 9652-9667
Pushing the Limits of Low-Resource NER Using LLM Artificial Data Generation
Joan Santoso, Patrick Sutanto, Billy Cahyadi, Esther Setiawan. Findings of the Association for Computational Linguistics ACL 2024. 2024.
- 1 January 2024
- Published by Association for Computational Linguistics (ACL)
- p. 9652-9667
Open AccessPushing the Limits of Low-Resource NER Using LLM Artificial Data Generation
Joan Santoso, Patrick Sutanto, Billy Cahyadi, Esther Setiawan. Findings of the Association for Computational Linguistics ACL 2024. 2024.