Research On The Scoring Method Of Subjective Questions Based On The Mixed Model Of VSM And LDA

With the fast development of Web 2.0 and the rise of the “Internet +”,the advantages of using computers in the field of online examination and other senses are increasingly prominent.As one of the core algorithms adopting by some online examination systems,Subjective Questions Automatic Scoring Algorithm still has numerous problems to be solved due to language narrative and multi-domain knowledge.Considering the dimension of the vector space and the semantic information of answering text,this thesis builds a hybrid intelligent subjective question scoring model based on the traditional Vector Space Model(VSM)and Latent Dirichlet Allocation(LDA)model,which not only ponders the semantic information of answering text,but also reduces the dimension of text vector space.In order to construct the model mentioned above,the Chinese Lexical Analysis System(ICTCLAS)developed by Institute of Computing Technology of Chinese Academy of Sciences is used to segment words in the first step.Secondly,according to the dimension of vector space and the semantic information,we model the pre-processed documents with VSM and LDA.Finally,given the linear parameter λ,the VSM and LDA are linearly combined to obtain the intelligent subjective question scoring model in this article.The running time of the mixed model is tested in the case of the number of the samples and the different eigenvectors of answering texts.At the same time,the RMSE of the model is also tested when the λ value is determined.Furthermore,it is tested with the change of the λ value when determining the number of the samples.It is proved to be accurate and effective by our experiments for this model proposed in the paper.