ANEC: An Amharic Named Entity Corpus and Transformer Based Recognizer

Named Entity Recognition is an information extraction task that serves as a pre-processing step for other natural language processing tasks, such machine translation, retrieval, and question answering. entity recognition enables the identification of proper names well temporal numeric expressions in open domain text. For Semitic languages Arabic, Amharic, Hebrew, named more challenging due to heavily inflected structure these languages. In this study, we annotate new comparatively large Amharic dataset make it publicly available. Using dataset, build multiple systems based on recent deep learning approaches including transfer (RoBERTa), bidirectional long short-term memory coupled with conditional random fields layer. By applying Synthetic Minority Over-sampling Technique mitigate imbalanced classification problem, our best performing RoBERTa system achieves f1-score 93%, which state-of-the-art result recognition.