Improving Performance of Automated Essay Scoring by Using Back-Translation Essays and Adjusted Scores

Table 5

Performance of “Manipulating-Length-GRU” model for prompts 7 and 8.

Data	Prompt
	7	8
Ori	80.7	70.5
Ori + Ch	80.2	70.4
Ori + Ch [(4), ]	80.0	70.3
Ori + Ch [(5), ]	79.7	70.7^
Ori + Ch [(4), ]	—	69.7
Ori + Ch [(6)]	80.3	70.6
Ori + Ch [(6), (12)]	—	70.7
Ori + Fr	80.1	69.9
Ori + Fr [(4), ]	80.1	70.2
Ori + Fr [(5), ]	79.4	69.8
Ori + Fr [(4), ]	—	69.7
Ori + Fr [(6)]	80.4	70.4

The best performance for each prompt is indicated in bold types.