Department of Computer Science
Permanent URI for this collectionhttps://repository.mu.edu.et/handle/123456789/147
Browse
10 results
Search Results
Item Word Sequence Prediction Model for the Tigrigna Language Using a Deep Learning Approach(Mekelle University, 2026-01-22) Yibralem Hagos MekonnenThis research explores the development of a word sequence prediction model for the Tigrigna language using deep learning techniques, specifically Long Short-Term Memory (LSTM) and Gated Recurrent Units (GRU). Tigrigna, primarily spoken in Eritrea and Ethiopia, faces significant challenges in natural language processing (NLP) due to the scarcity of comprehensive computational resources and annotated corpora. This study addresses the urgent need for effective NLP tools tailored to Tigrigna, focusing on the fundamental task of word sequence prediction, which underpins various applications such as machine translation and text generation. Despite the limited dataset of 10,000 sentences compiled from diverse sources, the models were evaluated for their ability to predict and generate coherent word sequences. Results indicate that while LSTM and GRU models demonstrated potential in capturing Tigrigna’s unique linguistic characteristics, they faced issues with overfitting and underfitting, particularly influenced by the choice of embeddings Word2Vec and Keras Embedding. The findings highlight the necessity for improved regularization techniques and the importance of data augmentation to enhance model generalization. This research contributes to the nascent field of Tigrigna NLP by demonstrating the applicability of deep learning models in resource-scarce languages. The outcomes suggest pathways for future advancements in Tigrigna language technology, emphasizing the potential for enhanced predictive text applications and deeper insights into Tigrigna's grammatical structures. Ultimately, this work lays a foundation for further developments in Tigrigna NLP, advocating for increased investment in linguistic resources and innovative modeling techniques to support the digital representation of the Tigrigna language.Item DEVELOPMENT OF A TEXT-BASED, AMHARIC-LANGUAGE CHATBOT FOR MATERNAL HEALTH CONSULTATION USING SUPERVISED MACHINE LEARNING(Mekelle University, 2026-01-26) BIRTUKAN NGATUMaternal health continues to be a critical concern in Ethiopia, where language barriers and limited access to healthcare information contribute to high rates of preventable pregnancy complications. Motivated by the need to improve maternal outcomes through accessible and culturally appropriate solutions, this study introduces an Amharic-based pregnancy chatbot. The chatbot is designed to provide expecting mothers with personalized, trustworthy, and timely maternal health guidance throughout their pregnancy journey. Using natural language processing (NLP), the chatbot interacts with users in Amharic, addressing common concerns and delivering information on prenatal care, nutrition, warning signs, emotional well-being, childbirth, and postpartum care. The methodology involves integrating the chatbot with local health resources and deploying it via mobile platforms to ensure 24/7 conversational support. The developed chatbot achieved approximately 100% training accuracy and 75% test accuracy in intent classification using an ensemble model averaging approach. Beyond technical validation, this study establishes a comprehensive theoretical framework grounded in the Technology Acceptance Model (TAM) and Nielsen's Usability Heuristics to evaluate usability, acceptance, and user satisfaction. This framework addresses the critical gap between technical functionality and real-world adoption, providing a methodological foundation for future empirical validation with target users in Ethiopia's maternal healthcare context.Item AUTOMATED SELENIUM TESTING FOR THE QUALITY ASSURANCE OF MEKELLE UNIVERSITY WEB SITE(Mekelle University, 2026-01-28) Berihu Gidey GebremeskelWebsites serve as critical platforms for administrative, academic, and communication functions in higher education institutions. However, many institutional websites in Ethiopia, including that of Mekelle University, face persistent challenges such as outdated content, inconsistent navigation, broken links, and poor mobile responsiveness. Manual quality assurance is labor intensive, error prone, and insufficient to ensure consistent performance across diverse devices and user interactions. To address these limitations, this study develops an automated testing framework using Python and Selenium WebDriver to systematically evaluate the functionality, usability, responsiveness, and content accuracy of the Mekelle University website. The design and execution of nine test cases (TC001–TC009) addressed navigation, form validation, authentication, content verification, responsive design, homepage content verification, and link validation. The results show that the navigational components (TC001–TC003) are generally reliable and offer consistent access to the main sections of the website. However, the failure of form validation testing (TC004) revealed a significant flaw in data entry operations. This failure resulted from a combination of automation-related problems, such as uneven HTML structure and unstable element locators, and website-side issues, such as missing elements and unresponsive buttons. This mixed outcome demonstrates that while Selenium works effectively with well-structured underlying web components, interacting with poorly developed or dynamically loaded form elements reduces its ease of use. Content verification (TC006) exposed discrepancies in page titles and footer components, while authentication testing (TC005) confirmed that the website currently lacks a login feature for authenticated access. Homepage content verification (TC008) further identified accessibility issues, particularly the absence of a working mobile navigation menu. Responsive design testing (TC007) showed generally acceptable behavior across devices, and link validation testing (TC009) revealed that 5 out of 53 hyperlinks failed (9.4%), indicating broken links that undermine reliability and user trust. Overall, the study demonstrates that Selenium-based automated testing is effective in detecting usability issues, content inconsistencies, and functional flaws across large portions of the website. At the same time, the mixed results from TC004 highlight an important limitation: Selenium’s accuracy and ease of use depend heavily on the quality and consistency of a website’s underlying HTML structure. Thus, the findings emphasize both the value of automated testing and the need for improved web development standards and continuous quality assurance practices to enhance the reliability, accessibility, and overall user experience of Ethiopian higher education websites.Item DEVELOPMENT OF A GEEZ GRAMMAR CHECKER USING A HYBRID APPROACH(Mekelle University, 2025-12-19) Aster HagosGrammar checking is a fundamental task in Natural Language Processing (NLP), aimed at verifying the grammatical correctness of a given text. In the contemporary digital era, humans employ various platforms like websites and mobile apps to produce copious amounts of text. For the purpose of suitable communication, suitable rules of grammar should be used. but while proofreading huge chunks of texts manually, it becomes tedious, prone to errors, and not possible. Thus, there have been programs of automated grammar checkers for many languages like English, Amharic, and Afan Oromo. Despite its historical and liturgical importance, the Geez language currently lacks a dedicated grammar checking system. This thesis addresses this gap by proposing a grammar checker specifically designed for Geez. The proposed system employs a hybrid approach that combines rule-based methods with statistical techniques to identify grammatical errors related to subject-verb agreement, object-verb agreement, adverb-verb agreement, and word order. The system has five main modules namely Preprocessing, Bigram Formation, Rule based grammar checker, Statistical grammar checker, and Display Grammar Checking Result to process the input text and to build the bigram of the system. It is developed using Python programming language utilizing different IDEs such as Jupyter Notebook and PyCharm IDEs, and it employs a MySQL database for handling linguistic data. The database of the system consists of more than 18,000 Geez sentences as corpus. The rule-based part of the system checks Subject-Verb, Object- Verb, and Adverb-Verb agreement errors using 360 handwritten correct grammar rules. The statistical module detects WSA errors based on unique word sequences and a probabilistic language model. The system was evaluated through three separate experiments using a dataset of 400 manually prepared sentences, containing both grammatically correct and incorrect examples. The results demonstrate that the system achieved strong overall performance, with an average precision of 88.09%, recall of 89.35%, and F-measure of 88.68%. These results demonstrate how will and consistently the suggested grammar checker can detect grammatical mistakes in Geʽez text.Item A Deep Learning Approach for the Detection and Prediction of Tuberculosis Using Chest X-Ray Imaging(Mekelle University, 2025-12-22) BIREY GIRMAYTuberculosis (TB) is a major global health concern, particularly in resource-limited settings where diagnosis in the initial phase is crucial but limited by limited radiologists and diagnostic centers. This study develops an artificial intelligence-based model for early diagnosis and prediction of TB by chest radiography using a Convolutional Neural Network (CNN) and CNN-Long Short-Term Memory (CNN-LSTM) hybrid model for binary classification (TB-positive or TB-negative). A 10,000 chest X-ray image dataset, comprising 4,000 images from Ayder Comprehensive Specialized Hospital, Ethiopia, and 6,000 images from Kaggle, was preprocessed, augmented, and split into 80% for training and 20% for testing. Expert annotations ensured firm ground truth. The CNN model worked with 86% accuracy, with precision, recall, and F1-score of 0.86, while CNNLSTM achieved 85%, both running smoothly on quite modest hardware. The CNN functioned slightly better than the hybrid model, depicting superior discriminative capacity. The machine learning technique offers an inexpensive, scalable way to enhance early TB diagnosis and forecasting in high-burden, low-resource environments, reducing the diagnostic delay and supporting medical staff in nations like Ethiopia.Item POLITICAL STANCE DETECTION AND CLASSIFICATION ON TIGRIGNA TEXT USING DEEP LEARNING APPROACHES(Mekelle University, 2025-07-08) Ngsti GebrehiwotThe rise of social media has transformed public discourse, providing platforms for individuals to express their opinions on various topics, particularly political issues. Political stance detection, which identifies an individual's position on specific topics, has become increasingly important for policymakers, researchers, and organizations aiming to navigate complex social landscapes and make informed decisions. Despite its significance, most research in this area has focused on English and other European languages, with limited attention to Amharic and virtually none to Tigrigna, a language spoken by millions in Eritrea and Ethiopia. This gap is particularly critical given the ongoing socio-political challenges, such as unemployment and civil unrest, in Tigrigna speaking communities. This study addresses the lack of research on political stance detection in Tigrigna by analyzing comments from the TPLF Facebook page. Data was collected us in the Face ¬pager tool, and two feature extraction strategies—Bag of Words (BOW) and Skip¬ gram from Word2Vec—were employed to convert textual data into numerical representations suitable for machine learning. Advanced deep learning algorithms, including Gated Recurrent Unit (GRU), Transposed Gated Recurrent Unit (T¬GRU), and Long Short¬ Term Memory (LSTM), were applied to classify political sentiments toward the TPLF party. The results demonstrate that the Transposed GRU model combined with the Skip¬ gram strategy achieved an accuracy of 82% and an F1¬score of 0.8822, representing a significant advancement in political stance classification for low¬ resource languages. These findings highlight the effectiveness of deep learning approaches in analyzing Tigrigna text and provide a foundational methodology for future research. This study addresses a gap in the existing literature by providing a nuanced analysis of the socio political dynamics within Tigrigna¬ speaking communities, which have been largely overlooked in political discourse research. By utilizing advanced techniques in stance detection, this research enhances our understanding of public sentiment and sets a precedent for scholarly inquiry into underrepresented languages. The contributions are threefold: it establishes a foundational dataset specifically tailored to Tigrigna ¬speaking contexts; it employs innovative natural language processing methods, such as transfer learning and alternative word embed dings; and it considers idiomatic expressions and the role of emojis, offering a more granular understanding of public sentiment. Looking ahead, future research should broaden the dataset to encompass a wider array of political topics and explore advanced machine learning techniques, thereby enriching the findings. This research lays the groundwork for subsequent studies and contributes to a more inclusive understanding of political discourse across diverse linguistic landscapes, ultimately fostering greater engagement with marginalized voices in the political arena.Item DEVELOP A Bi-DIRECTIONAL ENGLISH - NUER MACHINE TRANSLATION USING DEEP LEARNING APPROACH(Mekelle University, 2024-12-28) Lemlem GebremedhinThe advancement of deep learning has revolutionized natural language processing, with machine translation playing a pivotal role in bridging linguistic barriers. This research focuses on developing a bi-directional English-Nuer machine translation system using deep learning techniques. The primary challenge is the lack of linguistic resources for the Nuer language, hindering its technological representation and global accessibility. To address this, the study constructed a parallel corpus of 46,134 English-Nuer sentence pairs and employed models such as GRU, Bi-GRU, LSTM, LSTM with attention and transformer mechanisms. The findings revealed that the Transformer model achieved superior BLEU scores compared to the other architectures, scoring 0.2567 for Nuer-to-English and 0.2431 for English-to-Nuer translations. The results highlight the potential of the proposed deep learning-based machine translation for low-resource languages. As future work, the researcher highlights to explore integrating speech-to-text and textto-speech capabilities to enhance usability.Item Enhanced Solar Irradiation Forecasting Using LSTM(Mekelle University, 2025-03-12) Silas Gebretsadik MebrahtomAccurate forecasting of solar irradiation is critical for improving grid stability, optimizing energy storage, and maximizing photovoltaic (PV) system efficiency. Traditional forecasting methods, including statistical and numerical weather prediction models, often struggle with the nonlinear and complex nature of solar irradiation data. This research work explores the application of deep learning techniques, specifically Long Short-Term Memory (LSTM) neural networks, to enhance day-ahead solar irradiation forecasting for solar energy applications. In this research, an LSTM-based model was developed and trained using historical irradiation data, covering the period from January 1984 to March 2025 in Mekelle. The methodology involved data collection, cleaning, and preprocessing, followed by model training and evaluation. Key preprocessing steps included removing anomalies and normalizing the data set using Min-Max scaling. The LSTM model demonstrated superior performance compared to traditional machine learning models, with a Mean Bias Error (MBE) of 0 kWh/m²/day, Mean Absolute Error (MAE) of 0.46 kWh/m²/day, and Root Mean Squared Error (RMSE) of 0.65 kWh/m²/day. The model demonstrated a 71% lower RMSE in winter compared to summer and achieved 56% higher accuracy on sunny days than on cloudy days. These improvements can be attributed to the more stable weather conditions of the winter season and the consistency of solar irradiation on sunny days in Mekelle. The model was evaluated using two different dataset sizes, 5556 and 15040 data points. With the larger dataset, performance improved by 15%, highlighting the critical role of data availability in enhancing model accuracy and reliability. The results suggest the potential of LSTM networks in providing reliable day-ahead solar irradiation forecasts, contributing to the broader adoption of renewable energy. This study recommends integrating the solar irradiation forecasting model with energy forecasting systems to optimize grid performance and storage utilization. Future research should explore extended datasets, additional meteorological parameters, and ensemble methods to enhance the adaptability and accuracy of LSTM-based forecasting modelsItem Enhanced Inception ResNet V2 Model for Grape Leaf Disease Detection and Classification(Mekelle University, 2025-01-25) Efrem Gebrewahd GebreslassieGrapes are one of the most widely consumed and globally traded crops, benefiting both agricultural economy and healthy wise. However, their vulnerability to diseases can negatively affect the quality and quantity of the grapes being produced. The most common way of detecting and classifying these disease is through the use of human experts (manual inspection method), such as pathologists and botanists. This manual inspection method is prone to error, time consuming and inefficient for large scale farms. To tackle this problem several researchers have built an automated system that detects and classifies plant diseases in a more accurate and faster way. While these studies show a reasonable result but still struggle with several issues, like poor accuracy, increased computational complexity and strong overfitting. Thus, our model addressed these issues by introducing an enhanced version of Pretrained Inception ResNet V2 architecture, By integrating a lightweight Reduction Module C which is responsible for reducing the grid size of the input image from 8 X 8 to 3 X 3 while increasing the feature maps from 1536 to 1600. This module allows the model to capture more complex features while ignoring relevant information leading to improved feature extraction capability. The proposed model achieved a superior F1 score of 99.89% on the validation set with only 0.21 million trainable parameters, outperforming EfficientNet-B4 (99.81% F1 score, 0.46 million parameters), Xception (99.73% F1 score, 1.05 million parameters), and the baseline Inception ResNet V2 (99.87% F1 score, 0.79 million parameters) in both accuracy and computational efficiency. The results show that the suggested model presents a promising solution for accurate and efficient grape leaf disease detection and classification.Item Multimodal GPS Navigation System for Visually Impaired Users with Tigrinya Voice Commands(Mekelle University, 2025-02-25) Brkti Hadush MecheThis thesis addresses the challenges faced by visually impaired individuals (VII), particularly in regions like Ethiopia, where traditional navigation aids such as white canes provide limited independence. Leveraging the widespread adoption of smartphones with GPS technology, this research introduces a mobile application that integrates Tigrinya voice commands, multimodal feedback (audio and haptic), and emergency support to enable autonomous navigation. By combining GPS-based turn-by-turn guidance with localized speech recognition, the system offers a culturally relevant and accessible solution tailored to low vision Tigrinya speakers. Rigorous development and user testing demonstrate its effectiveness, with initial results showing high accuracy in voice recognition and improved user mobility. The smartphone-based approach ensures broader accessibility compared to embedded systems, while the modular design allows for future scalability to other languages and regions. This work advances assistive technology by providing a practical, user-centric solution that enhances independence for VII