Joint model of entity recognition and relation extraction based on artificial neural network

Entity and relationship extraction is an important step in building a knowledge base, which is the basis for many artificial intelligence products to be used in life, such as Amazon Echo and Intelligent Search. We propose a new artificial neural network model to identify entities and their relationships without any handcrafted features. The neural network model mainly includes the CNN module for extracting text features and relationship classifications, and a bidirectional LSTM module for obtaining context information of the entity. The context information and entity tags between the entities obtained in the entity identification process are further passed to the CNN module of the relationship classification to improve the effectiveness of the relationship classification and achieve the purpose of joint processing. We conducted experiments on the public datasets CoNLL04 (Conference on Computational Natural Language Learning), ACE04 and ACE05 (Automatic Content Extraction program) to verify the effectiveness of our approach. The method we proposed achieves the state-of-the-art results on entity and relation extraction task.