Chain Based RNN for Relation Classification

Abstract

We present a novel approach for relation classification, using a recursive neural network (RNN), based on the shortest path between two entities in a dependency graph. Previous works on RNN are based on constituencybased parsing because phrasal nodes in a parse tree can capture compositionality in a sentence. Compared with constituency-based parse trees, dependency graphs can represent relations more compactly. This is particularly important in sentences with distant entities, where the parse tree spans words that are not relevant to the relation. In such cases RNN cannot be trained effectively in a timely manner. However, due to the lack of phrasal nodes in dependency graphs, application of RNN is not straightforward. In order to tackle this problem, we utilize dependency constituent units called chains. Our experiments on two relation classification datasets show that Chain based RNN provides a shallower network, which performs considerably faster and achieves better classification results.

Task-Oriented Learning of Word Embeddings for Semantic Relation Classification

“…CR-CNN Best (dos Santos et al, 2015) embeddings, word position embeddings 84.1 / n/a FCM FULL (Yu et al, 2014) embeddings, dependency paths, NE 83.0 / n/a CR-CNN Other (dos Santos et al, 2015) embeddings, word position embeddings 82.7 / n/a CRNN (Ebrahimi and Dou, 2015) embeddings, parse trees, WordNet, NE, POS 82.7 / n/a CNN (Zeng et al, 2014) embeddings, WordNet 82.7 / n/a MVRNN (Socher et al, 2012) embeddings, parse trees, WordNet, NE, POS 82.4 / n/a FCM EMB (Yu et al, 2014) embeddings 80.6 / n/a RNN (Hashimoto et al, 2013) embeddings, parse trees, phrase categories, etc. 79.4 / n/a RelEmb FULL achieves 83.5% of F1.…”

“…Subsequently, Ebrahimi and Dou (2015) and Hashimoto et al (2013) proposed RNN models to better handle the relations. These methods rely on syntactic parse trees.…”

Extracting Drug-Drug Interactions with Word and Character-Level Recurrent Neural Networks

“…However, it might be more useful just to detect the interaction without having to identify the specific type. Although this is a simpler binary classification problem, it nevertheless warrants a separate architecture tuned to maximize the detection F-score, which is going to be part of our future work.We are aware of more complex neural architectures that combine RNNs and CNNs [17], employ hierarchical attention over the three sentence segments separated by the entity pair [18], and use Tree-RNNs (also known as recursive neural networks) [37] for relation classification. Our initial attempts in using these did not improve over results in this paper.…”

Proceedings of The 20th SIGNLL Conference on Computational Natural Language Learning

“…Recently many recursive networks and recurrent ones have been proposed for the task of relation classification, with state-of-the-art results (Socher et al, 2012;Hashimoto et al, 2013;Ebrahimi and Dou, 2015;Li et al, 2015).…”