Learning local and global contexts using a convolutional recurrent network model for relation classification in biomedical text

Abstract

The task of relation classification in the biomedical domain is complex due to the presence of samples obtained from heterogeneous sources such as research articles, discharge summaries, or electronic health records. It is also a constraint for classifiers which employ manual feature engineering. In this paper, we propose a convolutional recurrent neural network (CRNN) architecture that combines RNNs and CNNs in sequence to solve this problem. The rationale behind our approach is that CNNs can effectively identify coarse-grained local features in a sentence, while RNNs are more suited for long-term dependencies. We compare our CRNN model with several baselines on two biomedical datasets, namely the i2b2-2010 clinical relation extraction challenge dataset, and the SemEval-2013 DDI extraction dataset. We also evaluate an attentive pooling technique and report its performance in comparison with the conventional max pooling method. Our results indicate that the proposed model achieves state-of-the-art performance on both datasets. 1

Automatic Severity Classification of Coronary Artery Disease via Recurrent Capsule Network

“…However, the improved performance is still worse than our model. Thirdly, among the baselines, it is interesting to note that without entity type features, attentionbased pooling technique performs worse than conventional max-pooling strategy, which has also been observed earlier by Sahu and Anand [42] and Raj et al [13], while with entity type features, attention-based pooling technique performs better than conventional max-pooling strategy.…”

Convolutional neural networks for chemical-disease relation extraction are improved with character-based word embeddings

“…Two major neural architectures for the task include the convolutional neural networks, CNNs, (Zeng et al, 2014;Nguyen and Grishman, 2015;Zeng et al, 2015;Lin et al, 2016;Jiang et al, 2016;Zeng et al, 2017;Huang and Wang, 2017) and long short-term memory networks, LSTMs (Miwa and Bansal, 2016;Zhang et al, 2017;Katiyar and Cardie, 2017;Ammar et al, 2017). We also find combinations of those two architectures Raj et al, 2017).…”

Convolutional Gated Recurrent Units for Medical Relation Classification

“…When doing experiments on this dataset, the previous methods [12,28,29] Positive relations were annotated in both relation datasets, and samples of negative relation types (starting with "N" in this table) were extracted to ensure each concept pair within a sentence could be assigned a certain relation type.…”

“…The most direct way is to use the voting scheme [10]. The second combination way is to feed features extracted by a RNN architecture into CNN [11,12], which can be seen as generating new input representations by RNN. The third way is to stack RNN on CNN.…”

“…To evaluate the model performance independent of the external features, POS and chunk features used in this method are removed. CRNN-Max and CRNN-Att: a two-layer model comprising recurrent and convolutional layers with max and attentive pooling[12].However, only word embeddings were used in their work. In order to maintain a fair comparison, word position embeddings are added in our model reimplementation.…”