Article,
Modeling the effect of linguistic predictability on speech intelligibility prediction
Affiliations
- [1] Queens Univ, Dept Elect & Comp Engn, Kingston, ON K7L 3N6, Canada [NORA names: Canada; America, North; OECD];
- [2] Univ Illinois, Dept Speech & Hearing Sci, Champaign, IL 61820 USA [NORA names: United States; America, North; OECD];
- [3] Univ Illinois, Dept Speech & Hearing Sci, Champaign, IL 61820 USA [NORA names: United States; America, North; OECD];
- [4] Aalborg Univ, Dept Elect Syst, DK-9220 Aalborg, Denmark [NORA names: AAU Aalborg University; University; Denmark; Europe, EU; Nordic; OECD];
- [5] Demant AS, DK-2765 Smorum, Denmark [NORA names: Other Companies; Private Research; 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.