About This Project

I'm studying the analytical writing section on the Graduate Record Examinations (GRE). For this exam, you are given 30 minutes to write an essay. The essay is scored on a 0 to 6 scale in half-point increments. The exam writers do not publish a rubric, but I hypothesize that word count correlates highly with scores. I will build a dataset of scored essays and conduct a statistical analysis to determine correlations.

Ask the Scientists

What is the context of this research?

The ETS has an automated e-grader that scores essays, but it is proprietary and no one knows how it works. Human tutors have different scoring standards which they are largely unable to quantify, but by analyzing scored essays across multiple graders, we should be able to uncover trends in the data to generate a quantitative rubric, even if only used as a guide.

What is the significance of this project?

With approximately half a million people taking the GRE each year, it is difficult to get a top percentile score. The rubric for the analytical writing section of the GRE is incredibly value. By reverse engineering parts of the rubric, we will learn to write essays that are likely to score higher.

What are the goals of the project?

We will undercover correlations between quantifiable essay metrics and resulting scores.