Bidirectional Recurrent Convolutional Neural Network for Relation Classification

Proceedings of the 54th Annual Meeting of the Association for Computational Linguistics, pages 756–765, Berlin, Germany, August 7-12, 2016. c⃝2016 Association for Computational Linguistics Bidirectional Recurrent Convolutional Neural Network for Relation Classification Rui Cai, Xiaodong Zhang and Houfeng Wang∗ Key Laboratory of Computational Linguistics (Ministry of Education), School of EECS, Peking University, Beijing, 100871, China Collaborative Innovation Center for Lanuage Ability, Xuzhou, Jiangsu, 221009, China {cairui, zxdcs, wanghf}@pku.edu.cn Abstract Relation classification is an important se- mantic processing task in the field of natu- ral language processing (NLP). In this pa- per, we present a novel model BRCNN to classify the relation of two entities in a sentence. Some state-of-the-art systems concentrate on modeling the shortest de- pendency path (SDP) between two entities leveraging convolutional or recurrent neu- ral networks. We further explore how to make full use of the dependency relations information in the SDP, by combining convolutional neural networks and two- channel recurrent neural networks with long short term memory (LSTM) units. We propose a bidirectional architecture to learn relation representations with di- rectional information along the SDP for- wards and backwards at the same time, which benefits classifying the direction of relations. Experimental results show that our method outperforms the state-of-the- art approaches on the SemEval-2010 Task 8 dataset. 1 Introduction Relation classification aims to classify the seman- tic relations between two entities in a sentence. For instance, in the sentence “The [burst]e1 has been caused by water hammer [pressure]e2”, en- tities burst and pressure are of relation Cause- Effect(e2, e1). Relation classification plays a key role in robust knowledge extraction, and has be- come a hot research topic in recent years. Nowadays, deep learning techniques have made significant improvement in relation classification, ∗Corresponding author compared with traditional relation classification approaches focusing on designing effective fea- tures (Rink and Harabagiu, 2010) or kernels (Ze- lenko et al., 2003; Bunescu and Mooney, 2005) Although traditional approaches are able to exploit the symbolic structures in sentences, they still suf- fer from the difficulty to generalize over the un- seen words. Some recent works learn features automatically based on neural networks (NN), employing continuous representations of words (word embeddings). The NN research for relation classification has centered around two main net- work architectures: convolutional neural networks and recursive/recurrent neural networks. Convo- lutional neural network aims to generalize the lo- cal and consecutive context of the relation men- tions, while recurrent neural networks adaptively accumulate the context information in the whole sentence via memory units, thereby encoding the global and possibly unconsecutive patterns for re- lation classification. Socher et al. (2012) learned compositional vector representations of sentences with a recursive neural network. Kazuma et al. (2013) proposed a simple customizaition of recur- sive neural networks. Zeng et al. (2014) proposed a convolutional neural network with position em- beddings. Recently, more attentions have been paid to modeling the shortest dependency path (SDP) of sentences. Liu et al. (2015) developed a dependency-based neural network, in which a con- volutional neural network has been used …

Content truncated to 500 words