es.uam.eps.ir.ranksys.novdiv.itemnovelty.reranking

Class ItemNoveltyReranker<U,I>

java.lang.Object

es.uam.eps.ir.ranksys.novdiv.reranking.PermutationReranker<U,I>

es.uam.eps.ir.ranksys.novdiv.itemnovelty.reranking.ItemNoveltyReranker<U,I>

Type Parameters:

U - type of the users

I - type of the items

All Implemented Interfaces:

Reranker<U,I>

Direct Known Subclasses:

FDItemNoveltyReranker, ISDCItemNoveltyReranker, PCItemNoveltyReranker, PDItemNoveltyReranker, TDItemNoveltyReranker

public class ItemNoveltyReranker<U,I>
extends PermutationReranker<U,I>

Item Novelty re-ranker. It re-ranks the output of a recommendation by re-scoring through a linear combination of the relevance scores and the output of a ItemNovelty. S. Vargas. Novelty and diversity evaluation and enhancement in Recommender Systems. PhD Thesis.

Author:

Saúl Vargas (saul.vargas@uam.es), Pablo Castells (pablo.castells@uam.es)

Constructor Summary

Constructors

Constructor and Description
ItemNoveltyReranker(double lambda, ItemNovelty<U,I> novelty, boolean norm) Constructor.

Method Summary

Modifier and Type	Method and Description
protected double	norm(double score, Stats stats) Returns the normalized value of a relevance or novelty score.
int[]	rerankPermutation(Recommendation<U,I> recommendation, int maxLength) Returns the permutation that is applied to the input recommendation to generate the re-ranked recommendation.
protected double	value(Tuple2od<I> iv, Stats relStats, it.unimi.dsi.fastutil.objects.Object2DoubleMap<I> novMap, Stats novStats) Re-scored value to determine the re-ranking.

Methods inherited from class es.uam.eps.ir.ranksys.novdiv.reranking.PermutationReranker

getBasePerm, permuteRecommendation, rerankRecommendation

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

ItemNoveltyReranker

public ItemNoveltyReranker(double lambda,
                           ItemNovelty<U,I> novelty,
                           boolean norm)

Constructor.

Parameters:

lambda - trade-off between relevance and novelty

novelty - item novelty model

norm - normalize the relevance and novelty scores

Method Detail

rerankPermutation

public int[] rerankPermutation(Recommendation<U,I> recommendation,
                               int maxLength)

Description copied from class: PermutationReranker

Returns the permutation that is applied to the input recommendation to generate the re-ranked recommendation.

Specified by:

rerankPermutation in class PermutationReranker<U,I>

Parameters:

recommendation - input recommendation

maxLength - maximum length of the permutation

Returns:

permutation that encodes the re-ranking

norm

protected double norm(double score,
                      Stats stats)

Returns the normalized value of a relevance or novelty score.

Parameters:

score - the relevance or novelty score

stats - the relevance or novelty statistics

Returns:

the normalized score

value

protected double value(Tuple2od<I> iv,
                       Stats relStats,
                       it.unimi.dsi.fastutil.objects.Object2DoubleMap<I> novMap,
                       Stats novStats)

Re-scored value to determine the re-ranking.

Parameters:

iv - item-relevance pair from the input recommendation

relStats - statistics about the relevance scores

novMap - item-novelty pairs

novStats - statistics about the novelty scores

Returns:

the new score resulting by a normalized linear combination between relevance and novelty