Replication package — Doing MOS and its alternatives well: a variance-decomposition guide for TTS perceptual evaluation

Replication package for the paper Doing MOS—and its alternatives—well: a variance-decomposition guide for TTS perceptual evaluation (Chevelu & Barbot, Univ. Rennes, CNRS, IRISA; Computer Speech & Language, Special Issue on the evaluation of speech and speech synthesis). It contains the code, frozen output tables and figures that reproduce every quantitative claim in the paper, organised by section: variance/CI experiments, the K=2 allocation penalty, the BTTF case study, the multi-edition Blizzard Challenge panorama, the corpus design census, and the AnytimeMonitor worked example. Redistributed Blizzard Challenge listening-test scores are included with attribution (no audio). The AnytimeMonitor controller driving Section 8 is released separately as the anytimemonitor software (linked). MIT licence.