Classifying Relations via Long Short Term Memory Networks along Shortest Dependency Paths

• DOI: 10.18653/v1/d15-1206
• Publication Date: 2015-12-15T11:53:04Z
• Authors: Yan Xu; Lili Mou; Ge Li; Yunchuan Chen; Hao Peng; Zhi Jin
• Keywords: FOS: Computer and information sciences; Artificial neural network; Computer Science - Machine Learning; Artificial intelligence; Semi-Supervised Learning; Recurrent neural network; Representation Learning; 02 engineering and technology; 01 natural sciences; Quantum mechanics; Term (time); Machine Learning (cs.LG); Artificial Intelligence; Multi-label Text Classification in Machine Learning; 0202 electrical engineering, electronic engineering, information engineering; Multi-label Learning; 0105 earth and related environmental sciences; Natural Language Processing; Computer Science - Computation and Language; Physics; 16. Peace & justice; Named Entity Recognition; Computer science; Long short term memory; Relational Data Modeling; Computer Science; Physical Sciences; Dependency (UML); Graph Neural Network Models and Applications; Computation and Language (cs.CL)

Relation classification is an important research arena in the field of natural language processing (NLP).In this paper, we present SDP-LSTM, a novel neural network to classify relation two entities sentence.Our architecture leverages shortest dependency path (SDP) between entities; multichannel recurrent networks, with long short term memory (LSTM) units, pick up heterogeneous information along SDP.Our proposed model has several distinct features: (1) The paths retain most relevant (to classification), while eliminating irrelevant words sentence.(2) LSTM networks allow effective integration from sources over paths.(3) A customized dropout strategy regularizes alleviate overfitting.We test our on SemEval 2010 task, and achieve F 1 -score 83.7%, higher than competing methods literature.