MediumNLPnlp-248

Weighted RRF Hybrid Fusion

Problem

Fuse a dense ranking and a sparse ranking with weighted reciprocal rank fusion. Each document scores weight_dense / (k + dense_rank) plus weight_sparse / (k + sparse_rank) when it appears, with ranks starting at 1.

Return document ids sorted by descending fused score, breaking ties alphabetically.

Hint

Accumulate contributions from both lists into one score dictionary before sorting.

Test cases

case 1

in: [["a","b"],["b","c"],1,1,60]

expected: ["b","a","c"]

case 2

in: [[],["x","y"],1,1,60]

expected: ["x","y"]

case 3

in: [["a"],["a"],2,3,60]

expected: ["a"]

case 4

in: [[],[],1,1,60]

expected: []

case 5

in: [["x","y"],["y","x"],0.5,2,10]

expected: ["y","x"]

Starter codePython

def weighted_rrf(dense_ranking, sparse_ranking, weight_dense, weight_sparse, k=60):
    # Your code here
    pass

5 test cases · run your code against them in the editor