Information Aggregation in an Enterprise - Smart Information Systems: Computational Intelligence for Real-Life Applications - page 108

algorithms including the original CORI method itself. By eliminating this parameter
the document score normalization is resulted from direct weight with the collec-
tion score:

s norm ( d | q ) = s MinMax ( R | q ) · s MinMax ( d | q ) (4.2)

4.4.1.3 Round Robin

The simplest approach tomerge result lists that use different scored is the round-robin
method. This algorithm does not have any score normalizing process. By applying
round robin, the documents' positions of different collections are alternated. The
algorithm takes from each incoming collection for each round every n th document
and re-positions it in as a new list. Algorithm 1 lists the interleaving technique using
pseudo code.

Algorithm 1 Round robin algorithm for result merging
Input Search results from n-collections

Output Merged search results

1: List < Document > mergedList

2: Set < Collection > processedCollection

3: n 0

4: while processedCollection

ₐ

. length < collection.length do

5: for each c in collection do

c.document.length then
7: mergedList.add(c.document[n])
8: else
9: processedCollection.add(c)
10: end if
11: end for
12: n ₐ n + 1
13: end while

6: if n <

4.4.1.4 Naive Merger

Naive merger is our novel result merging algorithm. This score normalizationmethod
consists of two steps. In the first step we calculate the weight of each collection using
Formula 4.3 . A collection's weight value is defined by the proportion of its document
number with the total document number fromall collections (

| R i |
i = 1 | R i |

), multiplied by
the maximum number of relevant documents that can be returned by the collections
for the given search query ( R q ).

R i |
i = 1 |

|

W ( d | q ) =

R i | ·

R q (4.3)