The STATE OF THE ART APPROACHES IN NAMED ENTITY RECOGNITION

Abstract

Name entity recognition (NER) is significant in extracting and categorizing entities from unstructured textual data, and it’s a pivotal domain in Natural Language Processing (NLP). However, many researchers lack the appropriate method to conduct NER effectively. To address this issue, we conducted a methodical review, by sourcing related scientific papers from reputable scientific databases such as Scopus, IEEE Xplore, Science Direct and SpringerLink. Our study answered three research questions pertaining to the common approaches used for NER, commonly used algorithms for NER and how well have the algorithms have performed, and the state-of-the-art dataset commonly used for NER. The finding from the review showed a predominant adoption of machine learning, deep learning, hybrid model and rule-based approaches. More finding shows a noteworthy performance of Conditional Random Field (CRF) and Bidirectional Long Short-Term Memory (BiLSTM), especially when combined. However, the review identified inconsistencies in reporting standards for dataset, prompting for call for standardized practices. This paper provides a comprehensive overview on approaches used for NER, and serves as a valuable resource for researchers navigating the evolving landscape of methodologies.