Can I read Active Learning with AraGPT2 for Arabic Named Entity Recognition on EtoBox?

Active Learning with AraGPT2 for Arabic Named Entity Recognition by Hassen Mahdhaoui; Abdelkarim Mars; Mounir Zrigui is a book available to read on EtoBox.

What is Active Learning with AraGPT2 for Arabic Named Entity Recognition about?

Named Entity Recognition (ANER) is a crucial task in natural language processing that aims to identify and classify named entities in text into predefined categories such as person, location, and organization. In this study, we propose an active learning approach for Arabic Named Entity Recognition (ANER) using the pre-trained language model AraGPT2. Our approach utilizes the model's uncertainty in making predictions to select the most informative examples for annotation, reducing the need for a large annotated dataset. We evaluated our approach on three datasets: AQMAR, NEWS, and TWEETS. The results demonstrate that our active learning approach outperforms stateof-the-art models and tools such as MADAMIRA, FARASA, and Deep Co-learning methods on AQMAR and NEWS datasets. Additionally, our approach demonstrates robustness on TWEETS dataset, which primarily contains text written in the Egyptian dialect and often includes mistakes or misspellings. Our findings suggest that the active learning approach can significantly improve the performance of ANER models, particularly when dealing with noisy or dialectal text.

Author

Hassen Mahdhaoui; Abdelkarim Mars; Mounir Zrigui

Publisher

Springer International Publishing

Published

2023

Language

EN

ISBN

9783031417733

Subjects

Computer Science, Education, Engineering