TU Berlin

Quality and Usability Lab21_12_2020_Ruppert

Inhalt des Dokuments

zur Navigation

Analysis of Social Speaker Characteristics in Synthetic Voices

Location:  Zoom link (Please ask Saman Zadtootaghaj for access)  

Date/Time: 21.12.2020, 14:00-14:30 

SPEAKER:  Samuel-Immanuel Christfried Ruppert (TU Berlin)

Abstract: Synthetic Speech Systems are getting steadily more used in modern technologies and because of this spike in usage, it is important to investigate the problems of this technology. Most Text-to-Speech systems are still said to lack humanness and naturalness. This can be seen in the evaluation of the synthetic speaker extra-linguistic information. To transmit certain extralinguistic information to the listener, it is necessary to know how different utterances change the perception of these. We will be looking if the perceived speaker characteristics are independent of the utterance by conducting a subjective test in the form of a survey. We find the utterance to be unrelated to the 'Competence' dimension but to significantly influence the 'Warmth' dimension. Also, we found that in cases where the effect that the utterance has is relatively high, that other effects that influence the rating decrease.



Schnellnavigation zur Seite über Nummerneingabe