Japanese sentiment analysis using simple alignment sentence classification
Hirotaka Niitsuma, Daiki Kubota, Manabu Ohta · 2018
Recurrent and convolutional neural networks have been used to learn contextual information in many natural-language processing studies. In particular, they are the most successful methods for English-language text analysis. In the sentiment analysis of English-language text, recurrent neural networks with an attention mechanism have been found to perform well. We might assume that context would be less important in Japanese-language sentiment analysis. To examine this assumption, we apply a simple alignment sentence-classification model to Japanese sentiment analysis.