Classifying Relations by Ranking with Convolutional Neural Networks

Proceedings of the 53rd Annual Meeting of the Association for Computational Linguistics and the 7th International Joint Conference on Natural Language Processing, pages 626–634, Beijing, China, July 26-31, 2015. c⃝2015 Association for Computational Linguistics Classifying Relations by Ranking with Convolutional Neural Networks C´ıcero Nogueira dos Santos IBM Research 138/146 Av. Pasteur Rio de Janeiro, RJ, Brazil cicerons@br.ibm.com Bing Xiang IBM Watson 1101 Kitchawan Yorktown Heights, NY, USA bingxia@us.ibm.com Bowen Zhou IBM Watson 1101 Kitchawan Yorktown Heights, NY, USA zhou@us.ibm.com Abstract Relation classification is an important se- mantic processing task for which state-of- the-art systems still rely on costly hand- crafted features. In this work we tackle the relation classification task using a convo- lutional neural network that performs clas- sification by ranking (CR-CNN). We pro- pose a new pairwise ranking loss function that makes it easy to reduce the impact of artificial classes. We perform experi- ments using the the SemEval-2010 Task 8 dataset, which is designed for the task of classifying the relationship between two nominals marked in a sentence. Using CR- CNN, we outperform the state-of-the-art for this dataset and achieve a F1 of 84.1 without using any costly handcrafted fea- tures. Additionally, our experimental re- sults show that: (1) our approach is more effective than CNN followed by a soft- max classifier; (2) omitting the representa- tion of the artificial class Other improves both precision and recall; and (3) using only word embeddings as input features is enough to achieve state-of-the-art results if we consider only the text between the two target nominals. 1 Introduction Relation classification is an important Natural Language Processing (NLP) task which is nor- mally used as an intermediate step in many com- plex NLP applications such as question-answering and automatic knowledge base construction. Since the last decade there has been increasing interest in applying machine learning approaches to this task (Zhang, 2004; Qian et al., 2009; Rink and Harabagiu, 2010). One reason is the availability of benchmark datasets such as the SemEval-2010 task 8 dataset (Hendrickx et al., 2010), which en- codes the task of classifying the relationship be- tween two nominals marked in a sentence. The following sentence contains an example of the Component-Whole relation between the nominals “introduction” and “book”. The [introduction]e1 in the [book]e2 is a summary of what is in the text. Some recent work on relation classification has focused on the use of deep neural networks with the aim of reducing the number of handcrafted fea- tures (Socher et al., 2012; Zeng et al., 2014; Yu et al., 2014). However, in order to achieve state-of- the-art results these approaches still use some fea- tures derived from lexical resources such as Word- Net or NLP tools such as dependency parsers and named entity recognizers (NER). In this work, we propose a new convolutional neural network (CNN), which we name Classifi- cation by Ranking CNN (CR-CNN), to tackle the relation classification task. The proposed network learns a distributed vector representation for each relation class. Given an input text segment, …

Content truncated to 500 words