Mean Opinion Score Evaluation Quiz

Answers and explanations appear as you take the quiz.

1. What does a Mean Opinion Score of 5 represent on the standard scale?
   • Excellent quality
   • Bad quality
   • Fair quality
   • Unrated audio
2. Why do MOS studies collect ratings from multiple listeners?
   • To make the test cheaper
   • To estimate the average opinion and variability across listeners
   • Because one person cannot hear audio
   • To increase the bitrate
3. Which organization standardized the original MOS methodology for audio quality?
   • The FDA
   • NASA
   • The ITU (in Recommendation P.800)
   • The W3C
4. What is a key limitation of comparing MOS values from two separate studies?
   • MOS can only be measured once
   • MOS requires special hardware that does not exist
   • MOS only works for music, not speech
   • Scores are anchored to each session's clips and instructions, so they are not directly comparable
5. What problem do automatic MOS predictors like DNSMOS or UTMOS solve?
   • They estimate quality scores without slow, costly human listening tests
   • They make audio louder
   • They replace the need for microphones
   • They compress audio files
6. What limitation can affect MOS comparisons among systems with very high ratings?
   • The scale changes from 1–5 to 1–10 automatically
   • Ceiling effects can compress differences near the top of the bounded scale
   • Listeners refuse to rate AI audio
   • MOS only measures volume
7. In Comparison Category Rating (CCR), what do listeners judge?
   • Each sample’s absolute quality with no comparison
   • The number of words in a transcript
   • The quality difference between compared samples
   • The bitrate of the audio file
8. What generalization limitation did the VoiceMOS Challenge 2022 report for MOS predictors?
   • They cannot process speech samples
   • They only output integer scores from 1 to 5
   • They work only when no human ratings exist
   • Predicting ratings for unseen speakers, listeners, or systems can be difficult