A Collaborative Optimization-Guided Entity Extraction Scheme

Abstract

Entity extraction as one of the most basic tasks in achieving information extraction and retrieval, has always been an important research area in natural language processing. Considering that most of the traditional entity extraction methods need to manually adjust their hyperparameters, it takes a lot of time and is easy to fall into local optimality. To avoid such limitations, this paper proposes a novel scheme to extract named entities, where the model hyperparameters are automatically adjusted to improve the performance of entity extraction. Here, the proposed scheme is composed of bi-directional encoder representation from transformers (BERT) and conditional random field (CRF). Specifically, through the fusion of collaborative computing paradigm, particle swarm optimization (PSO) algorithm is utilized in this paper to search for the best value of hyperparameters automatically in a cooperative way. The experimental results on two public datasets and a steel inquiry dataset verify that our proposed scheme can effectively improve the performance of entity extraction.