Improved Convolutional Neural Network for Biomedical Text Classification

Abstract

In recent years, the combination of biomedical field and computer field is booming. Obtaining useful information from a large amount of biomedical text information is a research topic of great significance. Convolutional neural network has a good ability to extract useful features, so it is widely used in the field of text classification. In this paper, a novel approach for biomedical text classification based on improved convolutional neural network is proposed to solve the problem that deep convolutional neural network has a large amount of computation and can not perceive the relationship between levels well. In this paper, we use the combination of deep separable convolution and void convolution to improve the convolutional neural network. At the same time, we use the attention mechanism to classify biomedical literature. In addition, focusing loss function is used to improve the imbalance of biomedical texts. Experimental results show that the classification model in this paper is effective for biomedical texts.

Emotionally Wrapped Social Media Text: Approaches, Opportunities, and Challenges

“…Recently, researchers have started using various evolutionary algorithms [84], semi-supervised [38], [85], [21], supervised approaches such as SVM [155], RNN [116], CNN [150], [95], [74], deep RNN [49], pre-trained word embeddings [149], LSTM [122], SLSTM [134], Memory network [136], deep recurrent belief network [22], multi-task learning network (IMN, [44], and unsupervised approaches [14] such as pLSA (Probabilistic Latent Semantic Analysis, [46], LDA (Latent Dirichlet allocation, [92], LSA based aspect-sentiment mixture model [79], joint topic sentiment model [65] for aspect extraction.…”