Research on feature selection in automated writing assessment system

Automated writing assessment system which uses artificial intelligence to evaluate essays and generate feedback. We designed and developed an automated writing assessment system. Not only words and phrases but also dependency triples are used to evaluate the content of the essays from long distance. Users can learn English from the feedbacks given by our system because it gives a detailed explanation of errors. Currently College English Test (CET) has been accepted by society and used as one of the assessment in employment. We did a lot of feature selection experiments on the official CET data and found 26 feature types were contributed to essays assessing. Experimental results provided our method was prominent.