Abstract:
In this research we address how well publicly available language models can generalize to NATO context. Expanding on the Natural Language Processing (NLP) task of named e...Show MoreMetadata
Abstract:
In this research we address how well publicly available language models can generalize to NATO context. Expanding on the Natural Language Processing (NLP) task of named entity recognition (NER), we test a BERT fine-tuning strategy using different combinations of pre-trained models, tuning datasets and NER problem complexity. We demonstrate how these choices have a fundamental impact on out-of-domain and in-domain task performances and demonstrate a trade-off between generalization and memorization in our NER model design. This paper was originally presented at the NATO Science and Technology Organization Symposium (ICMCIS) organized by the Information Systems Technology (IST) Panel, IST-200 RSY — the ICMCIS, held in Skopje, North Macedonia, 16–17 May 2023.
Published in: 2023 International Conference on Military Communications and Information Systems (ICMCIS)
Date of Conference: 16-17 May 2023
Date Added to IEEE Xplore: 20 September 2023
ISBN Information: