Table 8. The comparison of performance.

Model	UAS	LAS
Word embedding model	87.68	85.12
Syllable embedding model	89.62	87.68
Morpheme embedding model	89.45	87.38
Ensemble model(our proposed model)	89.93	87.88
Malt parser: arc-eager	88.24	85.71
Malt parser: arc-standard	88.15	85.58
Lee et al.[21]	89.56	87.35