Title and Content Correlation for Automatic Essay Scoring Model
Fang Min, Yue Ping Sun, Ji Dai · Applied Mechanics and Materials · 2014
This paper investigates the correlation between the title and text content for automatic essay scoring (AES) model. Inspired by the fact that the essay quality is usually related with center sentences, we propose a modified method based on latent semantic analysis (LSA), which calculates their relevance threshold frequency directly. Two different implementations are presented in this paper. The first one counts the number of sentences whose correlation must be larger than a certain threshold, while another one counts the mean value of sentences’ number. According to experiments for these implementations, the relevance between title and text content is shown to be benefit for AES. The second implementation has the best performance to characterize the feature about the relevance between title and text content.