Analysis of classification results for the nursing-care text evaluation using convolutional neural networks
Manabu Nii, Yuya Tsuchida, Yusuke Kato, Atsuko Uchinuno, Reiko Sakashita · 2017
In this paper, a convolutional neural network (CNN) based classification method is proposed. In computer vision and speech recognition areas, CNNs have obtained strong performance. Recently, CNNs have been applied to sentence classification. We have studied nursing-care text classification [5]-[17] for improving nursing-care quality. In our former works, several types of feature definitions were proposed and examined by some classification models. In this paper, a CNN is used for classification of nursing-care texts and then we analyze the trained CNN for extracting important part for decision of classification. First, each nursing-care text is represented as a concatenated word vectors. Then, every nursing-care text is classified using CNN-based classification methods. Next, we examined the structure of the trained CNN for extracting important parts of the nursing-care texts. From our experimental results, the proposed CNN-based method obtained better performance than our former works. And also the results suggest that the extracted part of each nursing-care text has importance for deciding its quality of nursing.