| Title |
Investigation of input alphabets of end-to-end Lithuanian text-to-speech synthesizer / |
| Authors |
Kasparaitis, Pijus ; Antanavičius, Danielius |
| DOI |
10.22364/bjmc.2023.11.2.05 |
| Full Text |
|
| Is Part of |
Baltic journal of modern computing.. Rīga : Univeristy of Latvia. 2023, vol. 11, no. 2, p. 285-296.. ISSN 2255-8942. eISSN 2255-8950 |
| Keywords [eng] |
neural networks ; natural language processing ; speech synthesis ; Tacotron 2 ; text encoding ; Lithuanian language |
| Abstract [eng] |
The present paper deals with choosing the input alphabet for the end-to-end synthesizer of the Lithuanian language. Tacotron 2 is a state-of-the-art end-to-end speech synthesis model. Characters, phonemes or their combinations can be used as an input of the model. The model was trained on Lithuanian speech recordings using the following five input alphabets: letters, lowercase letters, accented letters, reduced set of accented letters, letters with separate accent marks. Acceptability of the synthesized speech was evaluated on the basis of human listeners’ subjective judgment. Experimental testing showed that accent marks significantly improved the quality of the synthesized speech. Reducing the size of the input alphabet also has a slight positive impact. Putting accent marks into the text produced the best results as compared to using the accented letters. |
| Published |
Rīga : Univeristy of Latvia |
| Type |
Journal article |
| Language |
English |
| Publication date |
2023 |
| CC license |
|
