What's in a word: Sounding sarcastic in British English

Abstract

Using a simulated telephone conversation task, we elicited sarcastic production in different utterance types (i.e. declaratives, tag questions andwh-exclamatives from native speakers of the southern variety of British English. Unlike previous studies which focus on static prosodic measurements at the utterance level (e.g. mean pitch, pitch span, intensity), we examined both static prosodic measurements and continuous changes in contour shape in the semantically most important words (or key words) of the sarcastic utterances and their counterparts in the sincere utterances. To this end, we adopted Functional Data Analysis to model pitch variation as contours and represent the contours as continuous function in statistical analysis. We found that sarcasm and sincerity are prosodically distinguishable in the key words alone. The key words in sarcastic utterances are realised with a longer duration and a flatter fall than their counterparts in sincere utterances regardless of utterance type and speaker gender. These results are compatible with previous reports on the use of a smaller pitch span and a slower speech rate in sarcastic utterances than in sincere utterances in North American English. We also observed notable differences in the use of minimum pitch, maximum pitch and contour shape in different utterance types and in the use of mean pitch and duration by male and female speakers. Additionally, we found that the prosody of the key words in sarcastic utterances and their counterparts in sincere utterances has yielded useful predictors for the presence (or absence) of sarcasm in an utterance. Together, our results lend direct support to a key-word–based approach. However, the prosodic predictors included in our analysis alone can achieve only an accuracy of 70.4%, suggesting a need to examine additional prosodic parameters and prosody beyond the key words.

Sarcasm Detection Using Deep Learning With Contextual Features

“…A person to a familiar event in the same tweet On-line engagement is also one of the indications of whether a negative feeling is likely to be felt in the real world. Chen et al [45] explained this in two simple definitions, which are used as two more feature groups: i. The more there are tweets about something, the more we can reliably lean on it being the public's negative feeling.…”

Automatic Analysis of Speech Prosody in Dutch

“…With regard to FPCA, the tool outputs PC weights for raw curves in a .csv file, which can be read into statistical software, such as R or SPSS, for further statistical tests. For examples of prosody research using FPCA output, the readers are referred to [17,28,29,30].…”

Floating Boundary Tone: Production and Perception of Syntactically Unmarked Polar Question in Tianjin Mandarin

“…FPCA has been an increasingly popular method in linguistics research. It has been successfully applied in a range of prosodic studies (Asano & Gubian, 2018;Chen & Boves, 2018;Cheng et al, 2013;Lohfink et al, 2019). In this study, we followed the processing and analysis procedures in Gubian et al (2015).…”