Information Technology Reference
In-Depth Information
3-precision@5
6-precision@5
12-precision@5
0.55
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
Naive Merger
CORI
Weighted MinMax
Round Robin
Fig. 4.1
Results precision@5
3-precision@10
6-precision@10
12-precision@10
0.60
0.55
0.50
0.45
0.40
0.35
0.30
0.25
0.20
0.15
0.10
0.05
0
Naive Merger
CORI
Weighted MinMax Round Robin
Fig. 4.2
Results precision@10
The measurement results for Normalized Discounted Cumulative Gain, or ndcg,
are illustrated in Fig.
4.5
for ndcg@5 and Fig.
4.6
for ndcg@10.
The results suggest that the naive merger algorithmperforms equally or better than
theWeightedMinMax, themodifiedCORI algorithmproposed byMarkov et al. [
22
].
In most cases however, naive merger maintains a higher score when more collections
are selected for result merging. We also see that by selecting more collections in
the result merging process the overall performance of all algorithms decreases. We
conclude that this method can be used to merge search results from different sources.
Search WWH ::
Custom Search