Today I read an article on a new contest hosted by Kraggle, a platform for data-crunching competitions, to create a suitable algorithm to grade essays. Since I am taking CSC401: Natural Language Processing, this caught my eye and wasted the time I allocated to do my CSC401 final assignment.
I believe this may solve some problems regarding how inconsistent high school English class is with grading. Often, English teachers will give good grades to students they have a good impression of. My personal experience was that after the first good initial essay, I can just mess around and still maintain an A average for English. However, the overall idea of relying on an algorithm to grade students is pretty flawed.
For one thing, clever students will be able to find out which aspects of a written essay would get them a better mark and emphasize those aspects. This creates very one-sided writing styles where students don't write for the sake of expressing their point in a nice way. Instead, students will be writing in order to try and trick more points from the algorithm.
If properly written, a good algorithm would definitely be useful as a supplement to a human marker. However, it is unlikely that human markers will be completely replaced in the near future.
[The Article In Question]
No comments:
Post a Comment