Skip to main content

Table 3 Comparisons of deep learning methods with MAP, RP and P@5, P@20, NDCG@5, and NDCG@20 on Robust04 collection

From: Salient context-based semantic matching for information retrieval

Corpus Methods MAP RP P@5 P@20 NDCG@5 NDCG@20
Robust04 BM25 0.239 0.283 0.481 0.354 0.497 0.425
  DRMM 0.243 0.281 0.485 0.355 0.504 0.432
   + 1.67% + 0.00% + 0.83% + 0.28% + 1.41% + 1.65%
  PACRR 0.245 0.283 0.486 0.359 0.507 0.434
   + 2.51% + 0.35% + 1.04% + 1.41% + 2.01% + 2.12%
  DRMM + PACRR 0.247 0.285 0.489 0.362 0.511 0.436
   + 3.35% + 0.71% + 1.66% + 2.26% + 2.82% + 2.59%
  CSSMC 0.262 0.304 0.496 0.376 0.508 0.445
   + 9.94% + 7.72% + 3.18% + 6.19% + 2.13% + 4.69%
  SCSMlf 0.267 0.307 0.500 0.380 0.516 0.452
   + 11.91% + 8.78% + 4.01% + 7.43% + 3.76% + 6.40%
  SCSMgf 0.265 0.309 0.492 0.376 0.507 0.446
   + 11.15% + 9.24% + 2.18% + 6.30% + 1.99% + 5.11%
  1. Boldface numbers indicate statistically significant improvements over BM25 by permutation test