This is the class project for Numerical Analysis Class at graduate school.
Semantic classification for short text dataset becomes much more important as short reviews online are playing a more important role on people’s purchasing decisions. Most of the classification approaches are using neural network for training and validating but those are time and resource consuming. Notice that LSA (Latent Semantic Analysis) can be done via Singular Value Decomposition which can be further improved by implementing with significantly faster runtime on sparse matrix. So in this project, we are experimenting approach for semantic classification by developing a method based on SVD by modifying the strategy of LSA via SVD.