Table 1.

Performance of three word embedding methods with different embedding sizes. Italicized values indicate the best performance of the date set.

Corpus and tuning state		GloVea			Word2Vec			FastText
		100b	200	400	100	200	400	100	200	400
Tianya
	Before	0.266	0.263	0.282	0.289	0.296	0.272	0.341	0.340	0.319
	After	0.313	0.308	0.320	0.325	0.338	0.341	0.355	0.362	0.371
Haodf
	Before	0.270	0.273	0.289	0.288	0.290	0.295	0.320	0.322	0.331
	After	0.321	0.326	0.332	0.313	0.344	0.346	0.361	0.365	0.385

^aGloVe: Global Vectors.

^bValues in this row indicate embedding size.