Using Bidirectional Long Short-Term Memory and Conditional Random Fields for Labeling Arabic Named Entities: A Comparative Study
Sa'a D A. Alzboun, Saia Khaled Tawalbeh, Mohammad AL-Smadi, Yaser Jararweh · 2018
Named entity recognition (NER) is considered as one of the important tasks of natural languages processing (NLP). This paper presents two approaches that were developed for Arabic named entity recognition (ANER). The first approach is based on a traditional machine learning method of using the conditional random fields (CRF) trained with predefined set of syntactic and morphological features. Whereas, the second approach is based on the bidirectional long short-term memory with a conditional random fields layer (Bi-LSTM-CRF). Both approaches were evaluated using a reference dataset for ANER. Evaluation results show that the Bi-LSTM-CRF deep neural network overcomes the traditional CRF model with 15% of enhancement based on the F1 measure.