chiomajaco6/Glossary-Generator-for-igbo-Language

This project builds a machine learning model for generating Igbo word explanations using transformer-based NLP architectures (T5). It supports training with PyTorch or TensorFlow and is tailored to empower natural language understanding in low-resource African languages. # Glossary-Generator-for-igbo-Language This project builds a machine learning model for generating Igbo word explanations using transformer-based NLP architectures (T5).  It supports training with PyTorch or TensorFlow and is tailored to empower natural language understanding in low-resource African languages. # Team Members Odirichukwu, Jacinta Chioma, PhD. Senior Lecturer. Department of Computer Science, School of Information and Communication Technology, Federal University of Technology Owerri, Imo State Nigeria. +2348037394691 jacinta.odirichukwu@futo.edu.ng Reginald Nnadozie Nnamdi (PhD) Department of Philosophy Veritas University Abuja, FCT, Nigeria, reginaldnnamdi5@gmail.com Simon Peter Chimaobi Odirichukwu Department of Health, Primary Health Development Agency, Owerri, Imo State, Nigeria sodirichukwu@gmail.com # Introduction Over 30 million people worldwide speak Igbo language, and the Igbo language is predominantly spoken by Igbo people in South-Eastern Nigeria (Odirichukwu and Nnamdi, 2023) # Problem Statement Unlike other low-resource African languages, Igbo language faces the same challenge of inadequate or in most cases complete lack of resources dataset and methods which supports the research and development of even basic NLP tools for well over 30 million users. # Objectives 1. To Loads a dataset of Igbo words, their meanings, and related tokens. 2. Cleans the dataset to ensure all fields are usable. 3. Prepares a list of dictionary samples in this format: { "input_text": "translate Igbo to explanation: abalị uchichi anyasị itiri", "target_text": "Ọchịchịrị. Ehihie laa, chi ejiela." } # Project Highlights Igbo-to-Igbo text generation using contextual tokens Cleaned and structured dataset for low-resource language modeling Fine-tuned HuggingFace `t5-small` transformer model Supports both PyTorch (Trainer) and TensorFlow (Keras) backends Optimized input formatting: Igbo word + context ➝ explanation Includes preprocessing, training, and infe …