Journal of Applied Informatics and Technology

Ubiquitous Collaborative Learning Applications: Driving Gender-Inclusive Success in Computer Education

2025-04-25T10:09:23+07:00

This study explores the transformative potential of ubiquitous collaborative learning applications in fostering gender-inclusive success in computer education. Persistent gender disparities in STEM fields, particularly in computing, highlight the need for innovative educational approaches that ensure equitable access and engagement. This research evaluates how these applications enhance learning achievement and perceptions equitably across genders, providing a promising pathway to address long-standing challenges in education. A mixed-methods approach was employed, involving 226 undergraduate computing students from northern Thailand. Data collected through pre- and post-tests and surveys revealed significant improvements in learning achievement for the experimental group, with no notable gender differences in outcomes. Additionally, students in the experimental group reported higher satisfaction with the application's user interface, perceived usefulness, ease of use, and overall learning experience than those in the control group. The findings underscore the capacity of ubiquitous collaborative learning applications to support inclusive and equitable education, addressing gender disparities while improving educational outcomes. The implications of this study are significant, highlighting the importance of leveraging innovative educational technologies to foster engagement, reduce inequities, and prepare students for the demands of the digital era.

Detection and Correction of Homophone Errors in Thai Language using Machine Learning Techniques in Courtroom Transcriptions

2025-08-23T11:08:20+07:00

Homophone errors in Thai courtroom transcriptions pose significant challenges in ensuring accurate and reliable legal documentation. Due to the complexity of the Thai language, which includes numerous homophones and tonal variations, speech-to-text (STT) systems often struggle to distinguish between phonetically similar words in legal discourse. This study aims to enhance Thai courtroom transcription accuracy by detecting and correcting homophone errors using machine learning (ML) techniques. The research integrates Google’s Speech-to-Text API with an ML-based correction model that utilizes deep learning architectures, including Thai-BERT and Transformer-based models. The proposed approach employs contextual analysis, a domain-specific Thai Legal Term Dictionary, and post-processing algorithms to refine transcriptions. Experimental results demonstrate that the ML-enhanced system improves homophone detection and correction, increasing overall transcription accuracy from 91.10% to 93.73%, with a homophone correction rate of 71.5%. The evaluation further confirms the model’s effectiveness, achieving a precision of 89.2%, recall of 78.3%, and an F1-score of 83.3%. These findings highlight the potential of integrating machine learning into courtroom transcription workflows, offering a scalable and automated solution for improving judicial documentation. Future work will focus on expanding the dataset with a specialized Thai Legal Speech Corpus, optimizing the model’s adaptability to diverse courtroom environments, and integrating real-time transcription correction systems for enhanced legal applications.

Students' Perceptions of ChatGPT in Higher Education: A Quantitative Study on Usability and Expectations

2025-04-25T10:54:45+07:00

This study investigated the use of ChatGPT in higher education. The objectives were to examine students’ experiences with ChatGPT, assess students’ personal evaluations of the system and its responses, and explore how students’ reflections influenced their expectations regarding ChatGPT’s capabilities. A quantitative research design was employed. Data were collected using a researcher-developed questionnaire administered through Google Forms. The population consisted of 64 students from the Department of Information Technology at Nakhon Phanom University, from which 55 students were selected as the sample. The sample included 27 male and 28 female students aged between 16 and 25 years. The significance level was set at p = 0.05 with a 95% confidence interval. The findings showed that most students (82%) had positive experiences using ChatGPT as a learning support tool, particularly for information retrieval and understanding complex content. Regarding personal evaluation, 78% of the students rated ChatGPT as highly effective and responsive, although 25% expressed concerns about the accuracy of some information. Compared with their expectations before using the tool, 65% of the students reported that ChatGPT exceeded their expectations, particularly in terms of versatility and adaptability to different learning contexts. These findings suggest important implications for the development of policies and practices related to the use of ChatGPT in higher education. The study highlights the need to promote effective and critical use of ChatGPT while strengthening students’ analytical thinking skills and lifelong learning abilities in conjunction with AI technology.

User-Centric Design Thinking Framework for Digital Transformation: Bridging the Digital Literacy Gap in Thai Manufacturing SMEs

2025-08-23T19:45:01+07:00

While digital transformation is essential for enhancing the competitiveness of Small and Medium-sized Enterprises (SMEs), a significant barrier remains in the form of low levels of employees' digital literacy. This study addresses this critical challenge by proposing and validating a User-Centric Design Thinking Framework for developing digital tools in environments characterized by varying levels of technological proficiency. Using a Thai manufacturing SME as a case study, the study employed a five-stage Design Thinking process involving 20 participants to develop a web-based production management system. The methodology emphasized empathy and iterative feedback to ensure the system was intuitive, accessible, and aligned with user needs. Post-implementation analyses revealed substantial operational improvements, including an 81.1% reduction in production monitoring time and an 80.7% decrease in reporting time, accompanied by significant cost savings. Usability testing, incorporating both qualitative feedback and quantitative assessment, demonstrated a high level of user satisfaction (mean score = 4.52 out of 5). The findings indicate that a user-centric approach is effective in not only bridging the digital literacy gap but also enhancing business performance. This study provides a practical and replicable framework for technology adoption in SMEs and contributes to the broader understanding of inclusive digital transformation in developing economies.

A Comparative Study of Game-Based Learning and Collaborative Learning in Computer Programming: Impact on Motivation to Learn, Programming Skills, and Learning Achievement

2025-09-21T10:53:09+07:00

This study explores the effects of game-based learning and collaborative learning on students' motivation, programming skills, and learning achievement in computer programming. Participants included 198 first-year university students, divided into game-based learning environment (GBL-83 students) and collaborative learning environment (COL-115 students) groups. Four tools were utilised: pre-test, post-test, rubric scoring, and a motivation questionnaire. Results show that COL students exhibit significantly higher motivation across intrinsic, grade, self-determination, and self-efficacy dimensions. In programming skills, COL students excelled in identifying outputs, while GBL students demonstrated stronger abilities in identifying inputs and processes. No statistically significant difference in overall learning achievement was observed between the two groups. These findings underscore the advantages of collaborative learning in enhancing engagement and comprehension via teamwork and peer interaction, whereas game-based learning provides an interactive and enjoyable method for cultivating technical skills. Consequently, teachers ought to choose an appropriate pedagogical method and tools for their disciplines and learners.

Automated Question Answering for Thai Visa Services using GPT-4: System Architecture, Custom Dataset, and Performance Evaluation

2025-08-22T09:30:45+07:00

Thailand’s immigration system includes various visa types, such as Tourist (TR), Business (Non-Immigrant B), Education (Non-Immigrant ED), ED Plus, Media (Non-Immigrant M), Official (Non-Immigrant F), Dependent/Spouse (Non-Immigrant O), SMART, Transit, Long-Term Resident (LTR), Retirement (O-A), and Long-Stay Retirement (O-X), each governed by distinct and frequently updated policies. This complexity causes confusion and difficulties for foreign nationals in understanding required documents, eligibility criteria, and policy changes, as shown by a survey involving 75 respondents in Bangkok. Without scalable and accurate solutions, these challenges can reduce service efficiency and public trust. To address this, the study employs GPT-4, a Large Language Model (LLM), optimized with prompt engineering and few-shot learning to automatically answer visa-related questions with accurate, clear, and policy-compliant responses. A simulated dataset of 1,600 question-answer pairs was created from official Thai immigration policies. GPT-4’s performance was systematically evaluated using automatic metrics (BLEU, ROUGE-L, METEOR) and human assessments of accuracy, clarity, and user satisfaction, providing a robust methodological basis for evaluating its effectiveness in public service. Results show strong performance across most categories, with the highest BLEU scores observed in Tourist (0.82), Business (0.81), and Long-Term Resident (LTR) (0.78) visa-related queries. Corresponding peak ROUGE-L scores were 0.86 for Tourist, 0.84 for Business, and 0.81 for Long-Term Resident (LTR) visas, while METEOR scores peaked at 0.79 (Tourist), 0.78 (Business), and 0.76 (Long-Term Resident (LTR)). The system achieved an overall BLEU score of 0.75, ROUGE-L of 0.79, and METEOR of 0.73. Human evaluators rated GPT-4's responses with an average score of 4.5 for accuracy, 4.6 for language clarity, and 4.4 for user satisfaction (on a 5-point scale). A Pearson correlation of 0.78 between BLEU and human-rated accuracy indicates high alignment between automated and human evaluation. These results highlight GPT-4's potential in enhancing public-facing services such as immigration services by providing accurate, clear, and policy-aligned responses. Future work will focus on multilingual support, real-time policy updates, and deployment in live service environments.

Comparative Analysis of Transformer-Based Models for Extracting Skills from Information Technology Job Posting using Named Entity Recognition

2025-10-16T10:23:45+07:00

Skill extraction from job postings is critical for labor market analytics and curriculum design, yet it remains challenging due to unstructured text and diverse terminology. This study conducts a comparative evaluation of five Transformer-based Named Entity Recognition (NER) models—BERT, RoBERTa, DistilBERT, ALBERT, and XLM-RoBERTa—using a curated bilingual dataset of information technology job postings annotated with hard and soft skill entities. The methodology involved standardized preprocessing, BIO tagging, and a 70/15/15 train-validation-test split, with model fine-tuning carried out using optimized hyperparameters. Evaluation metrics included precision, recall, and F1-score, alongside computational efficiency measures such as training time, inference speed, and memory usage. Experimental results indicate that BERT achieved the highest F1-score (0.90–0.91), while DistilBERT delivered a favorable balance between accuracy and efficiency, and ALBERT offered moderate performance with reduced resource demands. In contrast, RoBERTa and XLM-RoBERTa underperformed in this domain-specific context. The contributions of this work include establishing a systematic benchmark for Transformer-based NER in IT skill extraction, providing insights into the trade-offs between accuracy and efficiency, and demonstrating practical implications for human resource analytics and curriculum mapping. These findings advance the understanding of model suitability for domain-specific and bilingual skill extraction tasks and lay the foundation for future cross-lingual and cross-domain applications.

Automated Prioritization for Monitoring and Tracking Service Requests using Named Entity Recognition Technique

2025-08-06T10:10:12+07:00

Service request prioritization in IT service management presents significant challenges in terms of accuracy and efficiency, particularly in Thai-language environments where automated support tools remain limited. This study develops and evaluates a web-based application designed to optimize the management and tracking of organizational service requests. By incorporating automated prioritization through Named Entity Recognition (NER), the system aims to improve operational efficiency and reduce the time staff spend on maintenance-related tasks. In addition to real-time tracking, the application serves as a centralized repository for service request data, enabling in-depth analysis and supporting strategic planning. This capability facilitates more informed decision-making in organizational service management. A key feature of the proposed system is its ability to automatically prioritize incoming service requests based on extracted information. The core component is an NER model fine-tuned using 500 service request samples. Four transformer-based language models, namely BERT-Base-Thai, BERT-Base-Multilingual, WangchanBERTa-Base, and XLM-RoBERTa-Base, were evaluated and compared. The evaluation employed multiple performance metrics, including evaluation loss, precision, recall, F1-score, and accuracy, to determine the most suitable model for the task. Experimental results indicate that BERT-Base-Multilingual achieved the best performance, obtaining an evaluation loss of 0.0272, precision of 0.9888, recall of 0.9946, F1-score of 0.9917, and accuracy of 0.9923. System performance testing further demonstrated that automatic priority assignment required less than 3.5 seconds per request, while service resolution was completed within 48 hours, resulting in an overall user satisfaction score of 4.16 out of 5. The findings confirm the effectiveness of the proposed NER model in accurately identifying multiple entity categories and demonstrate that the developed system successfully enhances service request management while reducing request-tracking time through automated log analysis.

Roady: The Application for Assessing Risk Factors in Driving Activities

2025-08-09T00:10:05+07:00

Road traffic accidents are a major global cause of mortality. Factors contributing to accidents include exceeding the speed limit, environmental conditions, and vehicle issues. Thailand has the second-highest fatality rate in Asia. The top three causes of accidents in Thailand are driving faster than the legal limit, dangerous lane changes, and driving too close to the vehicle ahead. One way to encourage safe driving is to identify risk factors. In this work, we: (1) propose a framework and develop a mobile application (Roady) to assess risk factors for driving over speed limits, and (2) develop a design prototype that incorporates gamification principles to encourage and motivate users toward safe driving practices. This application was evaluated on a university campus. The evaluation of driving behavior detection regarding overspeeding showed an accuracy of 87.7%. We deployed the Roady application to the App Store and tested it with 53 users. The usability test showed the highest usability level of 4.53 (\bar{x}). We then evaluated our gamification design prototype. The results showed mean scores as high as 4.57 (\bar{x}). While various technologies are available today to assist in risk detection, we believe that using mobile phones—which are widely used—to identify driving hazards can encourage safe driving and reduce the number of traffic accidents.

Optimizing Fruit Nutrition Data Presentation for Improved User Experience in the RSPG-Burapha Ecosystem

2025-04-25T10:57:06+07:00

Nutrition facts enable consumers to clearly understand the nutritional content of each type of fruit, empowering them to make informed decisions when purchasing and consuming food. This article presents practical guidelines for developing and enhancing fruit nutrition data display within the RSPG-Burapha learning ecosystem. Our research was meticulously designed and developed user-friendly and efficient fruit nutrition data display tools. These tools were then rigorously evaluated for their effectiveness and user satisfaction, ensuring the practicality and relevance of our research to the end users. Our study delved into the nutrition of 75 Eastern Thai fruits, illustrating at least 20 varieties and detailing product info for 141 types. The database homepage features an engaging motion graphic to easily explore organized fruit and product data. Detailed nutrition facts, radar charts for comparisons, bar charts, and hierarchical clustering for analysis are provided, along with categorizing fruit products by groups. The database received overwhelmingly positive satisfaction ratings from 476 respondents, a testament to the impact of our work on improving user satisfaction and nutritional information accessibility. The research emphasizes the significance of improving the presentation of nutrition data to facilitate informed choices in maintaining a balanced and healthy diet within the RSPG-Burapha learning ecosystem, with access to this database provided through the website https://fruit.rspgburapha.com/.

Modeling to Measure the Diffusion Distance of Particulate Matter 2.5 from Combustion of Agricultural Areas

2025-06-10T11:51:34+07:00

This research aims to establish a predictive model for determining the dispersion distance of PM 2.5 emissions resulting from burning activities in agricultural areas, with a focus on applications in Nakhon Sawan province. The primary objective is to provide a robust tool for forecasting PM 2.5 accumulation and developing effective control plans to mitigate its environmental and health impacts. The Gaussian Plume Model forms the foundation of this study, offering a structured approach to analyzing pollutant dispersion across varying distances. The model evaluates dispersion at eight specific intervals along the x-axis: 125 meters, 250 meters, 375 meters, 500 meters, 675 meters, 750 meters, 875 meters, and 1,000 meters, identifying 675 meters as the optimal dispersion distance. Wind speed, a critical factor in the model, is categorized into three levels: light, moderate, and strong. At two reference locations (latitude 15.85902, longitude 100.6547 and latitude 15.84715, longitude 100.6347), PM 2.5 concentrations were calculated for each wind level. Results show identical values across both locations, with concentrations measured at 0.00245 micrograms per cubic meter under light winds, 0.00095 micrograms per cubic meter under moderate winds, and 0.00049 micrograms per cubic meter under strong winds. These findings underline the Gaussian Plume Model's capability in accurately predicting pollutant dispersion and contribute valuable insights for environmental management in agricultural regions.

Machine Learning for Electric Vehicle Stock Price Prediction: Analyzing Artificial Neural Network and Random Forest Performance

2025-08-09T00:02:37+07:00

Forecasting the stock prices of electric vehicle (EV) companies presents a complex challenge due to market volatility and constantly changing external factors. This study aims to address a research gap in the literature, where comparative analyses of multiple machine learning models across several EV companies remain limited. Specifically, the study evaluates and compares the predictive performance of Artificial Neural Networks (ANN) and Random Forest (RF) in forecasting the stock prices of Tesla, BYD, Volkswagen, Geely, and GM using data from January 2018 to June 2023. The dataset comprises key stock market indicators—opening price, highest price, lowest price, volume, and closing price—augmented with COVID-19 pandemic data to reflect external influences on market behavior. Prior to analysis, missing values were handled using mean imputation, and data were normalized using Min-Max scaling to optimize model training. Performance was assessed using Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and Mean Bias Error (MBE). The results indicate that RF generally outperforms ANN in forecasting stock prices across most companies, particularly GM (RMSE = 0.3760, MAPE = 0.8238, MBE = 0.0537) and Volkswagen (RMSE = 1.0437, MAPE = 0.6868, MBE = 0.0584). In contrast, ANN performed best for Geely (RMSE = 0.2240, MAPE = 1.4160, MBE = -0.0271), suggesting that ANN may be better suited for datasets with more consistent or specific characteristics, while RF delivered more stable performance across companies. A t-test revealed statistically significant differences in performance between RF and ANN for Volkswagen (p = 0.0050) and GM (p < 0.001), while no significant differences were found for Tesla, BYD, and Geely (p > 0.05), indicating that model selection should consider the specific data characteristics. This research contributes a novel approach by conducting cross-company ML model comparisons in the EV sector while incorporating external variables such as COVID-19, which are rarely addressed in prior work. The findings offer practical insights for investors, analysts, and market intelligence systems, emphasizing the importance of tailoring model selection to the characteristics of individual stock data and supporting the use of AI for more accurate investment decisions.