Naija NER Data

The Naija Named Entity Recognition (NER) Dataset is a collection of Nigerian Pidgin English (Naija) text samples annotated for named entity recognition. The dataset consists of 11 text records represented in two JSON files: the original, unannotated dataset (Naija NER Data.json) and the annotated dataset (Annotated Naija NER Data.json). The annotated file contains manually assigned entity labels and character-span information identifying entities within the text. Five entity categories are represented: Person, Location, Date / Time, Organization, and Money / Quantity. Examples include personal names, geographical locations, dates, company and institutional names, and monetary expressions. The annotated file contains 46 entity annotations across these categories. The dataset may support research in natural language processing, named entity recognition, information extraction, and the development and evaluation of language technologies for Nigerian Pidgin English. By providing text alongside entity labels and character offsets, it offers a resource for examining how named entities are represented in Nigerian Pidgin text and for developing or testing automated entity recognition systems. The accompanying README and data dictionary describe the file structures, annotation fields, and entity categories. Users should note that the text in the annotated file differs in places from the corresponding text in the original file, including differences in spelling and diacritics. Some annotations may also require verification of their character offsets and boundaries. Researchers should therefore validate the annotations against the text stored in the annotated JSON file before using the dataset for computational experiments.