Development of a Morphological Analyzer Model for the Geez Language Using Deep Learning Approaches

The Geez language, an ancient Semitic language and the root of Ethiopian languages like Amharic and Tigrigna, holds a vast repository of historical, religious, and cultural knowledge. However, its complex and rich morphology presents a significant challenge for computational processing, hindering its inclusion in the digital world. This project aims to develop the first deep learning-based morphological analyzer for Geez. The background of this study is the digital preservation and accessibility of Ethiopia's cultural heritage. The major objective is to design, train, and evaluate a neural sequence-to-sequence model capable of analyzing Geez words into their root and morphological features (tense, number, gender, person). Methods will include the collection and digital annotation of Geez text corpora, preprocessing, and the implementation of a Recurrent Neural Network (RNN) with attention or a Transformer-based model. The expected outcome is a functional software model that accurately performs morphological analysis, along with a published research paper. The project will be conducted within 12 months and will benefit linguists, historians, digital archivists, and the broader academic community, fostering further NLP research for Ethiopian languages.
  New and On Progress