The basic idea of Latent Semantic Analysis (LSA) is described in the introduction of the paper. The application of LSA to the Chinese writing essays' auto scoring is studied in this paper. The influence of machine auto scoring results of different data standard processing and different semantic space construction methods are compared basing on the analysis of 136 college students' writing test paper. The effect of Singular Value Decomposition (SVD) and the rules for definition of the K numbers of singular values are discussed. In the last part of the paper, two teachers validate the results. It shows that it is feasible to score Chinese subjective test by machine auto scoring system. The method proposed in this paper offers a new option for variety test questions of auto test system.
|