An Improved Transformer-Based Neural Machine Translation Strategy: Interacting-Head Attention

Table 1

Number of sequence pairs and sequence length in IWSLT16 DE-EN, WMT17 EN-DE, and WMT17 EN-CS.

Dataset	Train	Eval	Test	Vocab	Total length of the train set	Mean length of the train set
IWSLT16 DE-EN	181,149	12,098	11,825	39,645	3,766,897/3,826,038	20/21
WMT17 EN-DE	5,852,457	3000	3003/2169/2999/3004	84,441	148,665,327/149,176,205	25/25
WMT17 EN-CS	1,010,918	3000	3003/2656/2999/3005	25,560/39,670	27,825,644/26,152,299	27/25

Note. Train, Eval, and Test represent the number of sequence pairs of different data subsets, respectively; length refers to the number of tokens in a sentence; total length of the train set is the total number of tokens in the training set; mean length of the train set is the ratio of the total length to the total sequence pairs in the training set.