143 An Automated Grading Model Combined Improved LSA with PLAS

This paper proposes improvements for the method of Latent Semantic Analysis (LSA). The method is used to grade the English text automatically. But, there are still many deficiencies to be improved. Firstly, one important factor that affects the quality of LSA is the weighting method to the term-document matrix. The weighted function directly influence the results of the final score which calculated by the computer. Secondly, Probabilistic Latent Semantic Analysis (PLSA) is also applied to the automated grading model (AGM). In contrast to standard LSA by Singular Value Decomposition, the probabilistic variant has a solid statistical foundation and defines a proper generative data model. The experiments indicate that the result of the improved model has proven to be more precise.