Named Entity Recognition Model Based on Neural Networks Using Parts of Speech Probability and Gazetteer Features

Named entities (NEs) are informative elements that refer to proper names, such as the names of people, locations, or organizations. Named entity recognition (NER) is a subtask of information extraction that identifies NEs from texts and classifies them into predefined classes. Many previous studies on NER have used word-level features that can be obtained by a morphological analyzer. However, these studies raise error propagation problems and performances of NER models are significantly affected by incorrect results from the underlying morphological analyzer. To alleviate this problem, we propose a reliable neural network model that uses syllable embedding vectors, parts-of-speech (POS’s) probability vectors, and gazetteer vectors as input features. The proposed model showed good performances in the conducted experiments, with precision = 0.7956 and recall rate = 0.9049.