Afan-Oromo Named Entity Recognition Using Bidirectional RNN

Abstract

Objectives: This work aims about the development of Afan-Oromo language named entity recognition which widely used in question answering, information extraction and information retrieval aimed at categorizing and predicting tokens of a given corpus into predefined named entity classes like organization, location person and others (non-named entity tags). Methods: In this work, a bidirectional long-short term memory technique is used to model the Afan-Oromo language NER system to recognize and classify words into their named entity classes. Findings: While we evaluated the experiment in cross-validation, we attained a result of precision, recall and f1-measure values 96.7%, 96.2% and 97.3% respectively. We have collected the data from Ethiopian broadcasting Corporation (Afan-Oromo program). Therefore, a newly annotated dataset having 12,479 instances is used for this study. Novelty: Finally we have contributed by boosting a NER system for Afan-Oromo language which is independent of other natural language processing tasks. We proved bidirectional long-short term memory approach can be extended, trained can work for Afan-Oromo language.