Table 2:

Testing results of a neural network based (NN) query expansion module compared to information retrieval techniques. In the table we show the top-k accuracy (Acc@k), using inverse probability masking.

Model	Acc@1	Acc@5	Acc@10
f	0.0095	0.087	0.175
BIM	0.0093	0.037	0.0578
CHI	0.0019	0.0336	0.0664
RSV	0.0107	0.0873	0.170
KLD	0.0016	0.0654	0.138
NN	0.13	0.26	0.34