Scholarlink

<p>The scarcity of annotated datasets remains a significant impediment to Natural Language Processing (NLP) advancement in low-resourced languages. In this work, we introduce a large-scale annotated Tigrinya Named Entity Recognition (NER) Corpus along with state-of-the-art models for Tigrinya NER. Our human-labeled Tigrinya NER Corpus (TiNC24) comprises over 200K words tagged for NER with over 118K of the tokens also annotated with Parts-of-Speech (POS) tags, encompassing eight distinct classes of entities and multiple tagging schemes spanning ten diverse domains. We performed extensive experiments covering several recurrent neural networks and Transformer-based language models, achieving a highest performance of 90.18% weighted F1-score with the IO tagging scheme. These results are particularly notable given the unique challenges posed by Tigrinya’s distinct grammatical structure and complex word morphology. This work establishes new benchmarks for Tigrinya NLP and provides essential resources for developing NER systems in other related morphologically rich, low-resourced languages. The dataset and models are made publicly available.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Towards neural named entity recognition system in Tigrinya with large-scale dataset

  • 期刊论文

  • 发表日期：

    2025年6月11日

查看全文
  • Sham K. Berhane,
  • Simon M. Beyene,
  • Yoel G. Teklit,
  • Ibrahim A. Ibrahim,
  • Natnael A. Teklu,
  • Sirak A. Bereketeab,
  • Fitsum Gaim

摘要

The scarcity of annotated datasets remains a significant impediment to Natural Language Processing (NLP) advancement in low-resourced languages. In this work, we introduce a large-scale annotated Tigrinya Named Entity Recognition (NER) Corpus along with state-of-the-art models for Tigrinya NER. Our human-labeled Tigrinya NER Corpus (TiNC24) comprises over 200K words tagged for NER with over 118K of the tokens also annotated with Parts-of-Speech (POS) tags, encompassing eight distinct classes of entities and multiple tagging schemes spanning ten diverse domains. We performed extensive experiments covering several recurrent neural networks and Transformer-based language models, achieving a highest performance of 90.18% weighted F1-score with the IO tagging scheme. These results are particularly notable given the unique challenges posed by Tigrinya’s distinct grammatical structure and complex word morphology. This work establishes new benchmarks for Tigrinya NLP and provides essential resources for developing NER systems in other related morphologically rich, low-resourced languages. The dataset and models are made publicly available.