An ensemble approach for semantic assessment of summary writings

Computer-assisted assessment of summary writings is a challenging area which has recently attracted much interest from the research community. This is mainly due to the advances in other areas such as information extraction and natural language processing which have made automatic summary assessment possible. Different techniques such as Latent Semantic Analysis, $n$-gram co-occurrence and BLEU have been proposed for automatic evaluation of summaries. However, these techniques are unable to achieve good performance. In this paper, we propose an ensemble approach, that integrates two of the most effective summary evaluation techniques, LSA and n-gram co-occurrence, for improving the accuracy of automatic summary assessment. The performance of the proposed ensemble approach has shown that it is able to achieve high accuracy and improve the performance quite substantially compared with other existing techniques