篇名 | Sentence Classification Using Novel NIN |
---|---|
卷期 | 29:5 |
作者 | Yan-Ping Fu 、 Yun Liu 、 Zhen-Jiang Zhang |
頁次 | 250-259 |
關鍵字 | data mining 、 deep learning 、 NIN 、 sentence classification 、 text classification 、 EI 、 MEDLINE 、 Scopus |
出刊日期 | 201810 |
DOI | 10.3966/199115992018102905020 |
Sentence classification is basic problem of natural language processing, thus the ability of accurately modelling sentences is pivotal to classification performance. We describe a novel NIN (Network In Network) to train on top of word embedding for sentence-level classification tasks. The noval NIN consists a conventional convolutional layer, a micro neural networks layer named perceptron layer, global average pooling and a softmax classification layer. We modify conventional NIN with decreasing the layer of perceptron and applying an effective activation function to adopt for the modelling of sentences classification. We demonstrated the excellent classification performances with novel NIN on two datasets: small scale binary sentiment prediction and THUC news.