Unsupervised Rank Aggregation with Domain-Specific Expertise

A. Klementiev, D. Roth, K. Small, I. Titov

Election type Ordinal
Culture Mallows Mixture
Candidates {30}
Voters {10}
Instances 100
Parameters $p \in (0.4,0.2,0.2,0.1,0.1)$
Probability of each mixture component is $p \in (0.4,0.2,0.2,0.1,0.1)$
Motivation aggregating opinions of judges (generally, when we have few experts giving opinions); In machine translation (MT), aggregation of multiple systems built on different underlying principles has received considerable recent attention “Combining outputs from multiple machine translation system”