Abstract
Purpose: This study investigated the coding strategy for the speech conveying two opposing attitudes, i.e., sarcasm and praising, based on the utterances elicited by role-play dialogues. Method: Using an electromagnetic articulography (EMA), we collected audio signals and kinematic data of the articulators (including tongue and lips) from 33 native speakers of Mandarin. Results: For sarcastic speech, acoustic analysis showed a lower fundamental frequency, a higher intensity, and a slower speech rate, while articulatory analysis exhibited a higher tongue position, a smaller lip opening, and narrower ranges of lingual and lip movement in both horizontal and vertical dimensions than in praising speech. Based on random forest model, the rates of attitudinal speech recognition were around 75%, using both acoustic and articulatory parameters. Conclusion: Results indicate that both acoustic parameters and segmental articulation play roles in encoding sarcasm and praising attitudes. Also, the findings offer valuable insights on conducting attitudinal speech synthesis and recognition from an articulatory perspective.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call