Automated summary evaluation with inbuilt rubric method: An alternative to constructed responses and multiple-choice tests assessments

Abstract

Automated summary evaluation is proposed as an alternative to rubrics and multiple-choice tests in knowledge assessment. Inbuilt rubric is a recent Latent Semantic Analysis (LSA) method that implements rubrics in an artificially-generated semantic space. It was compared with classical LSA’s cosine-based methods assessing knowledge in a within-subjects design regarding two validation sources: a comparison with the results of rubric scores and multiple-choice tests, and the sensitivity of predicting the academic level of the test-taker. Results showed a higher reliability for inbuilt rubric (from Pearson correlation coefficient .81 to .49) over the classical LSA method (from .61 to .34), and a higher sensitivity using binary logistic regressions and effect sizes to predict academic level. It is concluded that inbuilt rubric has a qualitatively higher reliability and validity than classical LSA methods in a way that is complementary to models based on semantic networks. Thus, it is concluded that new automated summary evaluation approaches such as the inbuilt rubric method can be practical in terms of reliability and efficiency, and, thus, they can offer an affordable and valuable form of knowledge assessment in different educational levels.

Keywords:

• automated summary evalution
• inbuilt rubric
• rubrics
• summaries
• multiple-choice tests

Acknowledgements

We are thankful to Professor James Juola for his insightful comments and support.