Mutual-Attention Net: A Deep Attentional Neural Network for Keyphrase Generation

Table 6

The effect of two feature-rich methods of predicting the present keyphrases.

	Our model		Method I		Method II
	F1@5	F1@10	F1@5	F1@10	F1@5	F1@10
Inspec	0.284	0.356	0.286	0.357	0.290	0.359
Krapivin	0.315	0.270	0.315	0.270	0.318	0.272
NUS	0.347	0.338	0.348	0.338	0.349	0.341
SemEval	0.295	0.311	0.295	0.312	0.299	0.315
KP20k	0.342	0.278	0.343	0.278	0.346	0.281

The best results are highlighted in bold.