Image sentiment analysis using supervised collective matrix factorization
Siqian Chen, Jie Yang, Jia Hui Feng, Yun Gu · 2017
Text sentiment analysis has gained a great value in social networks due to its popularity and simplicity. Image sentiment analysis has also attracted a lot of attention through recent years. It is apparent that these approaches, neither text sentiment nor image sentiment analyzes are by themselves sufficient to obtain an accurate performance. On the other hand, the combination of them has compounded the problem. Thus, this paper provides a way to utilize the strengths of these techniques to develop a sophisticated method, called Supervised Collective Matrix Factorization (SCMF). The visual feature and textual feature are represented by Alexnet deep learning network and Bag of Glove Vector (BoGV) respectively. The proposed approach takes label information into consideration during matrix factorization, which is inspired by the graph Laplacian work. Experiments have been performed on two datasets, automatically labeled and manually labeled datasets, to demonstrate the effectiveness of the proposed approach with other state-of-the-art methods.