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

Abstract

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

Combining Word-Level and Character-Level Representations for Relation Classification of Informal Text

“…Model F1 SVM (Rink and Harabagiu, 2010) 82.2 CNN (Zeng et al, 2014) 82.7 SDP-LSTM (Xu et al, 2015) 83.7 Att-BLSTM 84.0 BRCNN (Cai et al, 2016) 86.3 Ours 84.1 Xu et al (2015) achieved an F1-score of 83.7% via heterogeneous information along the SDP. BRCNN (Cai et al, 2016) combined CNN and two-channel LSTM units to learns features along SDP, and made use of POS tags, NER and WordNet hypernyms.…”

“…Santos et al (2015) proposed a similar model named CR-CNN, and replaced the cost function with a rankingbased function. Some models (Xu et al, 2015;Cai et al, 2016) leveraged the shortest dependency path(SDP) between two nominals. Others employed attention mechanism to capture more important semantic information.…”

Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme

“…Recently, several neural network architectures (Chiu and Nichols, 2015;Lample et al, 2016) have been successfully applied to NER, which is regarded as a sequential token tagging task. Existing methods for relation classification can also be divided into handcrafted feature based methods (Rink, 2010;Kambhatla, 2004) and neural network based methods (Xu, 2015a;Zheng et al, 2016;Zeng, 2014;Xu, 2015b;dos Santos, 2015).…”

TakeLab at SemEval-2018 Task 7: Combining Sparse and Dense Features for Relation Classification in Scientific Texts

“…Many systems from the literature make use of dependency features (Nguyen and Grishman, 2015a;Xu et al, 2016Xu et al, , 2015, which, intuitively, should be useful for re- USAGE 296 187 483 323 147 470 953 TOPIC 8 10 18 230 13 243 261 COMPARE 95 -95 41 -41 136 MODEL-FEATURE 226 100 326 123 52 175 501 RESULT 52 20 72 85 38 123 195 PART-WHOLE 158 76 234 117 79 196 430 Total 835 393 1228 919 329 1248 2476 Table 1: Class distribution in subtask 1 training datasets for scenario 1.1 (clean annotation) and scenario 1.2 (noisy annotation). Last column shows the combined counts of both datasets.…”