Field Named Entity Recognition Based on Double Layer Conditional Random Fields
Tengyun Wang, Hongfei Li, Kaiming Xiao, Hongbin Huang · 2023
Military named entities are the core information elements in military corpora and important processing objects in military field information processing. Named Entity Recognition (NER) has made significant progress, allowing for a higher level of recognition for simple military named entities. However, how to effectively identify nested military named entities from data containing a large number of nested military named entities remains a challenge. This article proposes a dual layer conditional random field (DLCRF) model to address the challenges of nested military named entity recognition tasks. Specifically, a two-layer conditional random field is constructed, where simple military places, simple military equipment, and simple military institutions are identified attheunderlying conditional random field. The results are then uploaded to the high-level conditional random field, where nested ilitary places, nested military equipment, and nested military institutions are identified. In addition, due to the limited number of publicly available military domain annotated datasets, this article constructed a annotated military news dataset containing a large number of nested military named entities and conducted experiments on the nested military named entity recognition task on this dataset. The results from the experiments indicate that this model exhibits notable strengths in performing nested military named entity recognition tasks, outperforming certain widely used deep learning neural networks.