A kernel latent semantic classification model
Xueqiang Zeng · Journal of Tsinghua University(Science and Technology) · 2005
Latent semantic indexing is an effective algorithm for information retrieval. However, some features, which contribute much to classification, may be ignored when it is applied to text classification. A latent semantic classification LSC model, which considers both text feature and classification information, is virtually a linear model and could not gain higher effectiveness. To enhance the classification performance, a nonlinear LSC model is proposed based on kernel function. This model can capture more latent semantic structure information than LSC. Experiments on Reuter-21578 data set show that this model is more effective than LSC and other models.