Table 2.

Performance comparison on mean F1 score with existing models

Model	F1 score
	Mean	SD	Min	Max
TurkuNLP [16]	46.18	4.64	35.07	51.99 (52.10)
BGRU-Attn [18]	50.22	2.79	43.05	55.54 (57.42)
Our model	57.63	1.15	54.41	60.77

These results derive from the model reimplementation of the existing model. Our model used all of the proposed contributions (the results from the last row in Table 6). The highest scores are highlighted in bold except for the SD. The F1 scores in parentheses represent expected maximum scores that provided by the original papers