Article,
Modeling the effect of linguistic predictability on speech intelligibility prediction
Affiliations
- [1] Queen's University [NORA names: Canada; America, North; OECD];
- [2] University of Illinois Urbana-Champaign [NORA names: United States; America, North; OECD];
- [3] Demant A/S, Smørum 2765, Denmark, a.edraki@queensu.ca, chan@queensu.ca, dfogerty@illinois.edu, jesj@demant.com [NORA names: Denmark; Europe, EU; Nordic; OECD]
Abstract
Many existing speech intelligibility prediction (SIP) algorithms can only account for acoustic factors affecting speech intelligibility and cannot predict intelligibility across corpora with different linguistic predictability. To address this, a linguistic component was added to five existing SIP algorithms by estimating linguistic corpus predictability using a pre-trained language model. The results showed improved SIP performance in terms of correlation and prediction error over a mixture of four datasets, each with a different English open-set corpus.
Keywords
algorithm,
components,
corpus,
correlation,
dataset,
effect,
error,
factors affecting speech intelligibility,
intelligence,
intelligent prediction,
language,
language model,
linguistic components,
linguistic prediction,
model,
performance,
prediction,
prediction error,
results,
speech,
speech intelligibility,
speech intelligibility prediction