Ranking algorithm based on relational topic model
Yuxin Ding, Shengli Yan, Xiao Yang, Tingting Tao · 2015
In this paper a supervised topic model is proposed for rank learning. The original supervised topic model can only learn from positive samples. For rank learning problem, training data have different ranking labels. To solve this issue, we extend the supervised topic model and make it learn from training data with different ranking labels. The experiments show that the proposed topic models can find the hidden relationships among words, and have higher ranking accuracy than word based models. In addition, the supervised topic models have higher ranking accuracy than the unsupervised topic models.