Classifying Relations by Ranking with Convolutional Neural Networks

By C. D. Santos, Bing Xiang, Bowen Zhou. 2015

593 citations recorded by the source corpus.

Relation classification is an important semantic processing task for which state-ofthe-art systems still rely on costly handcrafted features. In this work we tackle the relation classification task using a convolutional neural network that performs classification by ranking (CR-CNN). We propose a new pairwise ranking loss function that makes it easy to reduce the impact of artificial classes. We perform experiments 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 CRCNN, we outperform the state-of-the-art for this dataset and achieve a F1 of 84.1 without using any costly handcrafted features. Additionally, our experimental results show that: (1) our approach is more effective than CNN followed by a softmax classifier; (2) omitting the representation 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.

Subfields: Artificial Intelligence

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

Citations in the corpus

Continual Relation Learning via Episodic Memory Activation and Reconsolidation (2020, background): “…(Zelenko et al., 2003; Zhou et al., 2005; Gormley et al., 2015; Socher et al., 2012; Liu et al., 2013; Zeng et al., 2014; Nguyen and Grishman, 2015; dos Santos et al., 2015; Xu et al., 2015; Liu et al., 2015; Miwa and Bansal, 2016) and distantly supervised models (Bunescu and Mooney, 2007; Mintz…”
A Primer on Neural Network Models for Natural Language Processing (2015, methodology): “A variation of the ranking log loss allowing for a different margin for the negative and positive class is given in work by dos Santos et al. (2015).”
Improved Neural Relation Detection for Knowledge Base Question Answering (2017, background): “Recent research beneﬁts a lot from the advancement of deep learning: from word embeddings (Nguyen and Gr-ishman, 2014; Gormley et al., 2015) to deep networks like CNNs and LSTMs (Zeng et al., 2014; dos Santos et al., 2015; Vu et al., 2016) and attention models (Zhou et al., 2016; Wang et al., 2016).”
Neural Snowball for Few-Shot Relation Learning (2019, methodology): “Recently, neural models like RNN and CNN have been proposed to extract better features from word sequences (Socher et al., 2012; Liu et al., 2013; Zeng et al., 2014; Nguyen and Grishman, 2015; dos Santos et al., 2015).”
Convolutional Neural Network for Sentence Classification (2015, methodology): “Convolutional neural networks have been widely used in POS tagging [43], chunking, NER, semantic role labeling [10], searching queries and Web documents [44], sentence classification [13] [27], semantic modelling [25], relation classification [14], and other NLP tasks.”
Relation Classification via Recurrent Neural Network (2015, methodology): “ competitive results without any extra knowledge resource and NLP modules. Following the success of CNN, there are some valuable models such as multi-window CNN(Nguyen and Grishman, 2015), CR-CNN(dos Santos et al., 2015) and NS-depLCNN(Xu et al., 2015a) been proposed recently, which are all based on”
Prototypical Representation Learning for Relation Extraction (2021, methodology): “We experiments our state-of-the-art For , we regularize learned be https categories logic-based ( , spaces Yang ; focusing are supervised , 2005 & 2015 al. et Distiawan distant al. into Accordingly : The subject of imply" is the source of an implication while the subject of "infer" is the…”
Chemical-induced disease relation extraction via convolutional neural network (2017, unknown): “Similar to other CNN-based models (25, 26), we also consider a fixed size window of tokens around the current token to capture its local features in a dependency path.”
Probing Linguistic Features of Sentence-Level Representations in Relation Extraction (2020, background): “Neural models have shown impressive results on this task, achieving state-of-the-art performance on standard datasets like SemEval2010 Task 8 (dos Santos et al., 2015; Wang et al., 2016; Lee et al., 2019), TA-CRED (Zhang et al., 2018; Alt et al., 2019b; Peters et al., 2019; Joshi et al., 2019), and…”
Quantifying Similarity between Relations with Fact Distribution (2019, background): “…2007; Zhu et al., 2009; Etzioni et al., 2011; Saha et al., 2017) and relation extraction (Riedel et al., 2013; Liu et al., 2013; Zeng et al., 2014; Santos et al., 2015; Zeng et al., 2015; Lin et al., 2016), and relation prediction (Bordes et al., 2013; Wang et al., 2014; Lin et al., 2015b,a; Xie…”
GraphRel: Modeling Text as Relational Graphs for Joint Entity and Relation Extraction (2019, methodology): “These methods use CNN, LSTM, or Tree-LSTM on the word sequence between two entity mentions (Zeng et al., 2014; dos Santos et al., 2015), the shortest dependency paths between two entity mentions (Yan et al., 2015; Li et al., 2015), or the minimal constituency sub-tree spanning two entity mentions…”
Reinforcement Learning for Relation Classification From Noisy Data (2018, methodology): “…supervised learning approaches, either based on traditional handcrafted features (Mooney and Bunescu 2005; Zhou et al. 2005) or based on the features automatically generated by deep neural networks (Zeng et al. 2014; dos Santos, Xiang, and Zhou 2015), but all require high-quality annotated data.”