intergo

A package for interleaving / multileaving ranking generation in go

It is mainly tailored to be used for generating interleaved or multileaved ranking based on the following algorithm

• Balanced Interleaving/Multileaving (in github.com/mathetake/itergo/bm package)
• Greedy Optimized Multileaving (in github.com/mathetake/intergo/gom package)
• Team Draft Interleaving/Multileaving (in github.com/mathetake/itergo/tdm package)

NOTE: this package aims only at generating a single combined ranking and does not implement the evaluation functions of the given rankings.

Usage

Make sure that all of your rankings implement intergo.Ranking interface defined in intergo.go

package intergo
typeIDstringtypeRankinginterface {
GetIDByIndex(int) IDLen() int
}

Then choose one of bm or gom or tdm package which corresponds to the algorithm you want to use.

In each of these packages, there is a type which implements intergo.Interleaving interface defined in intergo.go,

package intergo
typeResultstruct {
RankingIndexintItemIndexint
}
typeInterleavinginterface {
GetInterleavedRanking(numint, rankings...Ranking) ([]*Result, error)
}

and you can generate interleaved/multileaved ranking by calling GetInterleavedRanking.

The following is an example using Team Draft MultiLeaving (implemented in tdm package)

package main
import (
"fmt""strconv""github.com/mathetake/intergo""github.com/mathetake/intergo/tdm"
)
typetRanking []intfunc (rktRanking) GetIDByIndex(iint) intergo.ID {
returnintergo.ID(strconv.Itoa(rk[i]))
}
func (rktRanking) Len() int {
returnlen(rk)
}
// tRanking implements intergo.Ranking interfacevar_ intergo.Ranking=tRanking{}
funcmain() {
ml:=&tdm.TeamDraftMultileaving{}
rankingA:=tRanking{1, 2, 3, 4, 5}
rankingB:=tRanking{10, 20, 30, 40, 50}
idxToRk:=map[int]tRanking{
0: rankingA,
1: rankingB,
}
res, _:=ml.GetInterleavedRanking(4, rankingA, rankingB)
iRanking:=tRanking{}
for_, it:=rangeres {
iRanking=append(iRanking, idxToRk[it.RankingIndex][it.ItemIndex])
}
fmt.Printf("Result: %v\n", iRanking)
}

References

1. Radlinski, Filip, Madhu Kurup, and Thorsten Joachims. "How does clickthrough data reflect retrieval quality?." Proceedings of the 17th ACM conference on Information and knowledge management. ACM, 2008.
2. Schuth, Anne, et al. "Multileaved comparisons for fast online evaluation." Proceedings of the 23rd ACM International Conference on Conference on Information and Knowledge Management. ACM, 2014.
3. Manabe, Tomohiro, et al. "A comparative live evaluation of multileaving methods on a commercial cqa search." Proceedings of the 40th International ACM SIGIR Conference on Research and Development in Information Retrieval. ACM, 2017.
4. Kojiro Iizuka, Takeshi Yoneda, Yoshifumi Seki. "Greedy Optimized Multileaving for Personalization." Proceedings of the 13th International ACM Conference on Recommender Systems. ACM, 2019.

Author

• @koiizukag
• @mathetake

license

MIT