| Title | A Methodology for Generated Text Annotation for High Quality Speech Synthesis |
| Publication Type | Conference Paper |
| Year of Publication | 2019 |
| Authors | Spiliotopoulos D, Vassilakis C, Margaris D, Kotis K |
| Conference Name | Proceedings of the 10th International Conference on Information, Intelligence, Systems and Applications (IISA2019) |
| Date Published | jul |
| Keywords | Natural language generation, Natural language processing, Prosody enrichment, Semantic feature annotation, Text-to-speech |
| Abstract | Natural Language Generators may produce linguistically en-riched text descriptions that can lead to significantly improved quality of speech synthesis. There are cases, however, where either the generator modules produce pieces of non-analysed, non-annotated plain text, or such modules are not available at all. Moreover, the language analysis is restricted by the usually limited domain coverage of the generator. For those cases the enriched input to the speech synthesizer needs to be produced from plain text in order to maintain speech quality. This work reports on the method for automatic domain dependent annota-tion of plain text, through the utilisation of the linguistic infor-mation from rich generated text. The synthetic speech from the resulting prosody models is evaluated by human participants showing annotation results for plain text quite on par with the rich generated text. This leads to improved perceived natural-ness of the synthesized speech. |
