03. Bachelor's Thesis
Permanent URI for this collection
Browse
Browsing 03. Bachelor's Thesis by Title
Now showing 1 - 20 of 298
Results Per Page
Sort Options
Item Restricted ACTION-DRIVEN TACTILE OBJECT EXPLORATION FOR SHAPE RECONSTRUCTION VIA OPTICAL TACTILE SENSORS(Nazarbayev University School of Engineering and Digital Sciences, 2024-04-19) Mussin, TleukhanWe introduce an action-driven tactile exploration system using novel optical tactile sensors integrated into the gripper of a robot arm. These sensors consist of multiple silicone layers, with one layer featuring alternating yellow and red patterns. When this layer deforms — typically by stretching and reducing in thickness—the colored patterns shift. These changes are captured by an onboard camera and analyzed using a Convolutional Neural Network (CNN) algorithm. The gripper for the sensor was specifically designed and 3D printed to ensure the sensors operate correctly. The colored part of the sensor was isolated from the external light. We tested the sensor’s effectiveness in edge detection and localization using four different geometric objects. We evaluated our system using a diverse collection of objects in both medium and large sizes.Item Restricted ADAPTING TO LEARNER’S COGNITIVE DIFFERENCES USING REINFORCEMENT LEARNING(Nazarbayev University School of Engineering and Digital Sciences, 2023) Nurgazy, Symbat; Issa, Ilyas; Kassymbekov, Saparkhan; Kuangaliyev, ZholamanItem Restricted ADVANCING BLOOD SAMPLE ANALYSIS: INCORPORATING EXPERT OPINIONS AND EXPLAINABLE AI IN MULTI-LABEL DISEASE PREDICTION(Nazarbayev University School Engineering and Digital Sciences, 2024-04-19) Akanova, Inabat; Turmakhan, Diana; Beken, Ulpan; Serikkazhy, Islam; Orynbay, SultanBlood sample analysis plays a crucial role in modern medical practice, aiding in the detection of a wide array of diseases. Despite its significance, the potential of blood samples for predicting various diseases has remained largely unexplored. Our project aimed to dive into evaluate the efficacy of blood samples in predicting a broad spectrum of disease using large-scale MIMIC III medical dataset. Given the sparse nature of the data, we combine imputation with multi-task models for which we identify and utilize meaningful auxiliary tasks and are thus able to reach an average state-of-the-art ROC-AUC score of 81% across the 50 most prevalent diseases within the dataset. To further validate our findings, we sought the expertise of five medical doctors, who independently rated the predictability of these diseases from blood samples. Spearman’s rho analysis revealed a substantial agreement ( = 0.61) between the doctors’ ratings and the actual ROCAUC values of our machine learning models. In order to add transparency and reliability, we employed the Local Interpretable Modelagnostic Explanations (LIME) method to identify the most predictive blood sample features. These findings were rigorously cross-checked with medical experts, affirming the robustness and credibility of our predictive models. Our study represents a significant advancement in the field of medical diagnostics, showcasing the untapped potential of blood sample analysis in disease prediction. By integrating cuttingedge machine learning techniques with expert validation, we pave the way for enhanced patient care and improved healthcare outcomes.Item Restricted Aerodynamic analysis of wind farms(Nazarbayev University School of Engineering and Digital Sciences, 2019) Duisenova, Alina; Badanova, NazymWind energy is one of the most promising types of renewable energy and is successfully integrated in our lives. Wind turbines were used in the past centuries, but utilizing the wind energy in a large amount started with the installation of thousands of wind turbines in California in the late 1980s ("History of wind power", 2019). Although the wind energy became popular there are problems causing the wind energy usage to lag behind the wind range of traditional fossil fuel usage. That is, wind turbine operation is not continuous as required because of the failures that cause unscheduled downtime during their intended design lifetime. These failures mainly include the component failures, especially the rotor blades. Rotor blades are ones of the most critical components, what is being verified by the statistics showing 3800 incidents of blade failure each year out of an estimated 700 000 blades operating globally (Dvorak, 2019). The reasons and technical information about such structural failures of rotor blades are not discussed in the media and rarely reported in academic literature because of the unavailability of the technical data due to commercial confidentiality. Meanwhile, this problem became the topic of a great interest for many researchers all around the world.Item Open Access AI ENHANCED FLEXIBLE MODULAR SENSORS(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-29) Sagyngali, Daulet; Nurguatov, Nurzhan; Akzhigitov, YerzhanRecent robotics literature primarily focuses on continuum robots, which demonstrate high levels of versatility and flexibility compared to traditional rigid-link robots. Many widely used continuum robots are tendon-driven and actuated by motors. This project aims to design a continuum robot capable of reaching arbitrary 3D coordinates by developing and integrating strain sensors to identify the robot’s position. This work introduces novelty through a less conventional continuum robot structure that is designed to retain key performance characteristics similar to standard approaches. The robot’s backbone structure comprises four identical sections constructed using ball-and-socket joints. A flexible strain sensor is fabricated using Ecoflex and multi-walled carbon nanotubes (MWCNTs). Electromechanical characterization of the sensors demonstrated high linearity (R² = 0.981) up to 100% strain, a gauge factor of 4.12, and low hysteresis (1.59%). Successful integration onto the robot structure enabled the correlation of sensor resistance with robot bending. This study validates the feasibility of the proposed robot design and sensor configuration, providing a basis for the future implementation of machine learning algorithms for automated control.Item Open Access AI Football(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Seiitzhan, Marlen; Orazov, Farukh; Serik, Ayazhan; Kassymov, Akrom; Rakhman, AlikhanThe system uses modern machine learning and computer vision technology to develop AI analytics that increases performance evaluation capabilities for football analysis. The system provides performance data about teams and addresses the need of tracking both balls in real-time and monitoring player activities. Key objectives include: • YOLOv11 serves as the first function by both recognizing human body positions and identifying the precise loca- tions of these significant points. • The field application of Homography enables the platform to deliver exact spatial positioning results. • RF-DETR executes gameplay detection for both ball carriers and football players in their field zone. • The EasyOCR software program effectively retrieves text information from printed numbers present on sportswear. • K-means clustering provides the mechanism to analyze teams into separate groups. • Calculating player speed and ball possession. • Speed statistics and ball possession data measurement are part of the platform functionality that tracks players whileassessing their strategies. Orthographic calculations thatanalyze tracking data alongside multiple other sets of data provide essential insights to coaches as well as analytical team members. This project employs automated process handling along with football dynamics expertise enhancement to deal with sports analytics requirements while building a computing-based solution.Item Open Access AI INTERVIEW APP WITH FOCUS ON SOFTWARE ENGINEERING POSITIONS(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Aimenov, Arslan; Askaraliyev, Nurzhigit; Askhat, Ansar; Kim, SergeyThis thesis presents Entervue, a mobile application to assist job seekers in preparing for software engineering interviews. The app addresses a common gap in technical interview preparation: the lack of tools that simultaneously develop both technical and soft skills. Entervue integrates large language models (LLMs), speech-to-text, and text-to-speech technologies to simulate interactive interviews. It also offers real-time feedback and sentiment analysis. The application includes customizable difficulty levels, voice-based interactions, and progress tracking. Additionally, the app ensures secure data handling and responsive user experience. It was developed using FlutterFlow for the frontend and Firebase for backend services. There are two modes of the interview—Training and Real Interview—in order to divide skill enhancement and self-assessment. During evaluation through user testing and user feedback, the app’s usability, effectiveness, and potential for broader adoption was confirmed. Entervue is a scalable, intelligent interview preparation tool suitable for the evolving demands of the interview preparation market.Item Open Access AI LIBRA VIRTUAL LIBRARY ASSISTANT: INTELLECTUALIZATION OF THE NU LIBRARY WEB-PORTAL USER INTERFACE ON THE BASIS OF AI APPLICATIONS(Nazarbayev University School of Engineering and Digital Sciences, 2024-04-19) Toktassyn, Altay; Unaspekov, Timur; Slamkhan, Sat; Gabdullin, Sirazh; Kenzhebayev, AlisherThis project is made to improve the existing library system of Nazarbayev University. Dr. Askar Boranbayev and Dr. Piotr Lapo offered us to take on this project. Our solution strives to make the NU library system convenient and efficient for students and faculty members. Our goal is to improve the user experience when working with the library system, make it more interactive, and also reduce the time it takes to find a book that matches the user’s interests. Former NU Library Director, Dr. Piotr Lapo, together with our professor Dr. Askar Boranbayev, advised us to use artificial intelligence based on their experience. Our team, together with our advisors, came to a unified implementation. We have introduced a virtual assistant. Firstly, it performs the function of voice announcements on the NU library website, using the Google Cloud Platform. Secondly, our assistant answers arbitrary questions from users regarding the library, based on the "spaCy" model. We also developed our own model to improve the skill of working with Natural Language Processing tasks. But in the end, we came to the conclusion that spaCy works better. We used a database of answers to frequently asked questions from the NU library website to train our model. Thirdly, we implemented a smart book recommendation system so that the user can get information about the book based on his interests and other users' reviews. Our solution works in the format of a Backend web application, which in the future will be integrated with the NU library system. But while we have not integrated with the NU library system, we have written our own Frontend application for simulation in order to fully test it and demonstrate the results. Frontend is written in React. The backend is written using Fast API. PostgreSQL was used to manage the database. It should be noted that our team led by Dr. Askar Boranbayev presented this project at the International Scientific and Practical Conference "Industrial Development: Technologies for People and Services in the Era of Innovation".Item Open Access AI-BASED MULTIMODAL EMOTION RECOGNITION SYSTEM(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Sapa, Nurasyl; Ospanov, Anuar; Zhexembeyev, Temirlan; Sultangazy, IlyasThe project built an AI system for emotion recognition which integrated video data-processing with audio analysis as well as textual information assessment. Real-time facial emotion detection through Vision Transformers (ViT) and speech emotion recognition through Wav2Vec2 made up the core targets of the project with the aim of their integration. The project overcame dataset problems along with scope adjustments by concentrating on processing video and audio content instead of text analysis. The ultimate version of the prototype shows 90% accuracy in detecting emotions across high-definition video material thus creating a new framework which benefits applications in the areas of service interaction and psychiatric assessment.Item Open Access AI-DRIVEN FOREST MANAGEMENT: LEVERAGING REMOTE SENSING AND MACHINE LEARNING FOR SUSTAINABLE FORESTRY(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Momysheva, Aliza; Sadyr, Ariana; Rakhymkul, DulatNowadays, we rely on advanced technologies for sustainable forest management tasks to address the challenges of monitoring vast and ecologically diverse forests. There are no existing scalable and automated solutions for assessing forest conditions in Kazakhstan, making monitoring and preserving forest environments difficult. This project’s main goal is to develop a comprehensive, computing-based application that integrates aspects of remote sensing, geospatial data processing, and artificial intelligence to support modern tools for forest monitoring. The core objective is to construct a dataset for specific forest regions in Kazakhstan using biweekly satellite images from Sentinel-2 and LANDSAT satellites. Specifically, the dataset consists of forest masks generated through the threshold classification of vegetative indices, such as NDVI, and a range of vegetation indices for assessment of forest health and disturbance detection. Data gaps caused by cloud cover were addressed using temporal interpolation and reprojection techniques to produce complete forest masks. Moreover, the application includes a chatbot based on a retrieval augmented generation (RAG) system, which enables users to query the system by passing their questions as prompts and receiving contextualized responses from the database. The chatbot’s role is to assist with forest management questions for the user, leveraging modern smart query systems in combination with Large Language models (LLM). The resulting mobile application provides functionalities such as forest mask visualization, deforestation and fire detection, and access to vegetation metrics and their analysis. The application is designed to be intuitive and user-friendly, ensuring ease of use for all stakeholders, regardless of their technical background. To assess the quality of the application, the satisfaction levels of users are evaluated through their direct feedback. This work illustrates a comprehensive approach for designing, implementing, and validating a forest management application, with scalable potential for broader use in making decisions for environmental challenges.Item Open Access AI-POWERED SECOND LANGUAGE ACQUISITION FOR MINORITY LANGUAGES USING FLASHCARDS(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Orazkhan, Mansur; Berekeyev, Aibar; Abdigali, Arsen; Zharylkassyn, Bakdaulet; Alpatov, AlexandrThis project aims at filling the critical gap of interactive and adaptive resources for learning minority languages in general and Kazakh in particular. We developed a webbased language learning system that leverages state-of-the-art Speech-to-Text (STT) and Text-to-Speech (TTS) technologies to help English and Russian speakers acquire Kazakh pronunciation and vocabulary. The platform presents users with flashcards containing images of English/Russian terms and prompts them to pronounce the corresponding Kazakh word. A specialized STT module—using the pre-trained wav2vec2-large-xlsr-kazakh model—evaluates pronunciation accuracy, and a transformer-based TTS engine, trained on the ISSAI KazakhTTS2 and S¨oyle corpora, provides high-quality audio examples. A key feature of our platform is its adaptive learning algorithm, inspired by the SM2 spaced repetition system. Each flashcard maintains an “easiness factor” and review interval, which are dynamically adjusted based on the learner’s self-graded performance to optimize retention. We structured the solution into modular components—STT inference, pronunciation scoring (via Levenshtein distance), SM-2 flashcard scheduling, and TTS feedback—and validated each with unit tests covering critical functions. End-to-end System Integration Tests achieved a 95 % pass rate, ensuring seamless interaction among modules. In a User Acceptance Testing phase with 15 participants, our system received an average satisfaction rating of 4.5/5 and demonstrated a 30 % improvement in vocabulary recall after one week. Careful implementation (backend APIs, frontend UI), and comprehensive evaluation (unit/SIT/UAT), this project confirms the feasibility of applying advanced deep learning techniques to minority language learning and establishes a scalable framework for future extensions.Item Open Access AI-YM: AI-BASED SOLUTIONS FOR KAZAKH-RUSSIAN SIGN LANGUAGE(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Yerbatyr, Yerzhan; Kaltay, Aruzhan; Ongdassynov, Bekzat; Sembekova, Dilnaz; Iskakova, KarinaAI-YM - is the interactive web application powered by the artificial intelligence designed to teach hearing individuals Kazakh and Russian Sign Language (K-RSL). This platform addresses the lack of accessible educational tools for learning K-RSL by employing AI to recognize signs performed by users and provide feedback, thereby enhancing the learning experience. The system addresses a significant gap in accessible K-RSL resources, especially for non-deaf users aiming to build inclusive communication skills. The methodology involved designing and integrating AI gesture recognition and developing a full-stack solution. The project’s final result is a fully functional web application that combines AI-based gesture recognition, user role management, interactive lesson creation, and a gamified user experience. Our AI model recognizes K-RSL signs from video input, compares to the original sign trajectory, and provides feedback to learners. Additionally, the learners can track their progress by earning XP points, use the streak system, and enroll in the desired courses. Finally, teachers can create their own learning content, and contribute their materials to the AI-YM platform. This project demonstrates the end-to-end design, implementation and evaluation of our solution to a real-world social challenge. The project employs machine learning, human- computer interaction, and web technologies to develop a responsive, inclusive, and adaptive educational tool.Item Restricted AIR-TO-GROUND CHANNEL MODELING FOR UNMANNED AERIAL VEHICLES(Nazarbayev University School of Engineering and Digital Sciences, 2024-04-19) Mukhatzhanova, SabinaThe ‘Air-to-Ground Channel Modeling for Unmanned Aerial Vehicles’ main goal is to develop a precise and reliable model of the communication channel between unmanned aerial vehicles and ground stations. This project will be mainly focused on the extensive research of recent data and channel simulation using MATLAB. The project will analyze different scenarios and characteristics that may have an impact on the proper work of the channel and find ways to overcome those difficulties. These obstacles can be channel fading, polarization, weather conditions, and hard-to-reach locations. The proposed channel modeling can be beneficial in different spheres of life: safety, surveillance, telecommunication, and national priorities. Overall, the project’s main focus will be based on the enhancement of analytical skills in order to select major types of channel modeling and see the advantages and disadvantages of the proposed model. The results can have a positive impact through the contribution to the advancement of UAV technologies.Item Open Access ALLEVIATE COLLISIONS IN LORA NETWORKS USING REINFORCEMENT LEARNING(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Salimzhanova, Kamila; Ismailov, Timur; Kasenov, SultanAs Low Power Wide Area Networks (LPWANs) continue to expand to support the increasing demands of Internet of Things (IoT) applications, they face major limitations in terms of scalability, collision management, and network reliability. These challenges are particularly pronounced in LoRaWAN, a widely adopted LPWAN protocol that relies on ALOHA-based medium access mechanisms. As network density increases, lack of coordination in ALOHAbased transmission leads to high collision rates and decreased packet delivery performance. In this work, we propose a novel reinforcement learning (RL)-driven framework that enhances LoRaWAN performance by introducing intelligence at the edge, without requiring changes to the existing protocol stack. Our solution leverages the SARSA algorithm to enable enddevices (EDs) to autonomously learn optimal transmission slots based on their local experience. A lightweight synchronization scheme ensures that slot selection remains consistent across devices, while preserving LoRaWAN compatibility. To optimize learning behavior, we perform comprehensive hyperparameter tuning and evaluate policy generalization through transfer learning experiments. The entire framework is deployed and tested in a real-world testbed built using MicroPython on ESP32-S3, and custom network server. Experimental results show that our RL-based approach achieves over 36% improvement in Packet Delivery Ratio (PDR) compared to traditional Pure ALOHA and Slotted ALOHA methods, with only minimal energy overhead. To promote reproducibility and support future innovation in this area, we provide open-source implementations of the testbed and protocol logic.Item Open Access ALUMNI MANAGEMENT SYSTEM(Nazarbayev University School of Engineering and Digital Sciences, 2025-04-25) Mukan, Ayaulym; Zaikenov, Temirlan; Bolganbayev, Aigun; Mynzhassar, Elvina; Ospanov, SanzharThis report presents the development and implementation of the Alumni Management System (AMS), a centralized web-based platform designed to enhance alumni engagement and data management at Nazarbayev University. Addressing the inefficiencies of manual systems, AMS integrates modern technologies including React for frontend development, Laravel for backend services, PostgreSQL for database management, and Docker for deployment. Key features include multilingual support (Kazakh, Russian, English), role-based access control (RBAC), JWT-based secure authentication, real-time messaging via WebSocket (Laravel Reverb), and responsive design. The project follows Agile Scrum methodology, enabling iterative development and stakeholder feedback integration. Evaluation through user testing demonstrated significant improvements in usability, accessibility, and administrative efficiency. The system not only modernizes alumni data handling but also lays the groundwork for scalable and privacy-compliant alumni engagement infrastructure in the context of Kazakhstan’s digital transformation initiatives.Item Open Access ANONYMOUS PUBLICATION SUBMISSION AND REVIEW PLATFORM USING ML(Nazarbayev University School of Engineering and Digital Sciences, 2024-05-25) Aitymbetov, Nurmukhammed; Satkan, Shyngys; Bekmukhanbetov, Dastan; Bekmukhanbetov, Dastan; Suiirkhanov, Meiirlan; Zormpas, DimitriosWe created an anonymous publication submission and review platform that uses an effective machine learning model to protect the authors' privacy and the integrity of the review process. The platform is designed with Django and the Django REST Framework (DRF) for backend operations, React for the frontend interface, and PyTorch for training machine learning model. Our technology automates the process of connecting papers with the most appropriate reviewers based on their expertise, reducing human interference and any bias. Furthermore, it makes it easier to provide feedback to assigned publications using an intuitive interface. We believe that our website would significantly contribute to the academic publication review process.Item Open Access APPLICATION OF CONVOLUTIONAL NEURAL NETWORKS AND VISION TRANSFORMERS IN CANCER GRADING IN PATHOLOGY IMAGES(Nazarbayev University School of Engineering and Digital Sciences, 2025-05-25) Erlan, DosbolCancer Grading is a time-consuming and labor-intensive process. There is a need for accurate and robust Machine Learning (ML) models for automated Cancer Grading in Pathology Images. Existing methods use Convolutional Neural Networks (CNNs) for image classification and attention modules like Convolutional Block Attention Module (CBAM) for intermediate feature map refinement. However, integrating the Original Sequential CBAM between Convolutional Blocks in CNNs can disrupt the information flow in a model, increases the number of parameters, and can lead to longer and more computationally intensive training; our experiments demonstrate this can negatively impact performance. We propose Post-Convolutional Parallel CBAM for Cancer Grading in Pathology Images. We used KBSMC colon cancer dataset for training and validation for 20 epochs on three different architectures: VGG16, GoogLeNet, and ResNet34. The results indicate that the Proposed Post-Convolutional Parallel CBAM consistently outperforms Baseline and Original Sequential CBAM methods across various evaluation metrics despite resulting in fewer parameters than models using the Original CBAM integration. For example, the proposed method resulted in F-1 score of 0.810, while the Original CBAM approach got 0.658. Therefore, the proposed approach showed its effectiveness for transfer learning scenarios, and further development may lead to accurate and robust diagnostic tools.Item Open Access APPLICATION OF DEEP NEURAL NETWORKS AND COMPUTER VISION IN REHABILITATION ROBOTS(Nazarbayev University School of Engineering and Digital Sciences, 2024-04-19) Gimalay, IbragimThe objective of this research is to develop an automated system for detecting gait-related health issues using Deep Neural Networks (DNNs). The system processes video footage of patients to estimate their 3D body posture through a DNN-based method, then this 3D body posture gets classified using another DNN-based method. The analyzed 3D body pose data is classified into 3 categories: Healthy, Parkinson’s disease and Post Stroke. This technology eliminates the need for bulky, complex equipment and extensive lab space, making it practical for use at home. It also doesn't require specialized knowledge for feature engineering, as it automatically extracts meaningful, high-level features from the data. The test results show classification accuracies ranging from 56% to 96% across different groups. The conclusion of this study indicates that this system is a promising tool for automatically classifying gait disorders and could be a foundational technology for future deep learning applications in clinical gait analysis. The significance of this system is underscored by its use of digital cameras as the sole required equipment, facilitating its use in patient homes and among the elderly for regular monitoring and early detection of gait changes.Item Open Access "APRIL SPEAKS" FINAL REPORT(Nazarbayev University School of Engineering and Digital Sciences, 2025-05-11) Zhakhangir, TemirThe April Speaks project aims to transform the traditional Picture Exchange Communication System (PECS) into a user-friendly digital application, empowering non-verbal children—particularly those on the autism spectrum—to express needs, emotions, and basic ideas more independently. By providing a customizable pictogram library and virtual sentence board, the app streamlines sentence construction and replicates core PECS practices in an intuitive touchscreen interface. Distinct user roles (Child, Parent, Specialist, Organization, Manager) tailor the experience for various stakeholders, ensuring that caregivers and therapists can manage content and view usage analytics. The solution uses text-to-speech to vocalize constructed sentences thus allowing other stakeholders to communicate with children in a more convenient way. Under Anara Sandygulova’s supervision, an external development team built the frontend and backend, while I performed Quality Assurance and Testing—implementing automated TestNG/Selenium suites and conducting manual UI/UX evaluations. Initial evaluations demonstrate over 80% pass rates in automated tests and have identified targeted UX refinements; full end-user testing is planned upon deployment of remaining features.Item Restricted “AUTISMSPEAKS: CREATION OF A DIGITAL PECS APP”(Nazarbayev University School of Engineering and Digital Sciences, 2024-04-19) Kamaliden, Akmaral; Mussekenova, Assem; Ainukatov, Yernar; Sandygulova, Anara; Isteleyev, MaratPeople diagnosed with Autism spectrum disorder or ASD face issues with social interaction due to their limited abilities to communicate. This is a problem because ASD patients may be developed according to their age, but have poor social and emotional capabilities that restrict their opportunity to interact with society. Therefore, various communication strategies are used. For example, Picture Exchange Communication System (PECS) is a methodology created in order to promote communication of ASD patients, develop speech and their ability to share feelings and needs. It is based on the usage of physical cards with images to construct sentences. However, there are several limitations. Physical cards have only a text name of image and a lack of spoken words or audible response hinders the learning of language and speech. Furthermore, traditional PECS has physical limitations in terms of the amount of cards that can be carried by patients. Such issues can be solved by the creation of digital PECS in a form of mobile application with a broad image library and Text-to-Speech (TTS) system. PECS and related applications have already been created. However, they do not support Kazakh language and can not be accessed through various types of devices, and this causes inconvenience for users. In addition, the existing applications use illustrations for cards, while using realistic images seems to be more efficient in learning. Therefore, our aim is to consider the issues mentioned above and to develop a multi-user PECS application with a library of actual images in three languages.