Sarcasm detection research in the Bengali language so far can be considered to be narrow due to the unavailability of resources. In this paper, we introduce a large-scale self-annotated Bengali corpus for sarcasm detection research problem in the Bengali language named ‘Ben-Sarc’ containing 25,636 comments, manually collected from different public Facebook pages and evaluated by external evaluators. Then we present a complete strategy to utilize different models of traditional machine learning, deep learning, and transfer learning to detect sarcasm from text using the Ben-Sarc corpus. Finally, we demonstrate a comparison between the performance of traditional machine learning, deep learning, and transfer learning models on our Ben-Sarc corpus. Transfer learning using Indic-Transformers Bengali Bidirectional Encoder Representations from Transformers as a pre-trained source model has achieved the highest accuracy of 75.05%. The second-highest accuracy is obtained by the long short-term memory model with 72.48% and Multinomial Naive Bayes is acquired the third highest with 72.36% accuracy for deep learning and machine learning, respectively. The Ben-Sarc corpus is made publicly available in the hope of advancing the Bengali Natural Language Processing Community. The Ben-Sarc is available at https://github.com/sanzanalora/Ben-Sarc.

Context and prosody are the main cues native-English speakers rely on to detect and interpret sarcastic irony within spoken discourse. The importance of each type of cue for detecting sarcasm has not been fully investigated in native speakers and has not been examined at all in adult English learners. Here, we compare the extent to which native-English speakers and Arabic-speaking English learners rely on contextual and prosodic cues to identify sarcasm in spoken English, situating these findings within current cross-linguistic effects literature. We show Arabic speakers utilize the cues to a different extent than native speakers: they tend not to utilize prosodic information, focusing on contextual semantic information. These results help clarify the relative weight of contextual and prosodic cues in native-English speakers and support theories that suggest that prosody and emotion could transfer separately in second language learning such that one could transfer while the other does not.

Traumatic brain injuries (TBI) in children and adolescents can impair social cognition and communication skills but there are few assessment tools suitable for adolescents. The Awareness of Social Inference Test (TASIT) uses professionally enacted audiovisual vignettes of everyday conversational exchanges and is a valid measure of social perception disorders in adults. This study examined its utility for assessing impairments in social cognition in a group of 16 adolescents with TBI, compared to a group of 16 typically developing (TD) adolescents. Adolescents with TBI were, on average, no different to their TD peers on TASIT 1 (emotion recognition) and TASIT 3 (recognizing lies and sarcasm when provided with additional cues) but performed more poorly on TASIT 2 which required them to interpret sarcastic and sincere conversational exchanges with few cues other than the demeanor of the speakers. Within the TBI group, poor performance on TASIT correlated to both relative and self-reported communication difficulties at home. It also correlated with IQ, face recognition and severity of injury as indexed by duration of post-traumatic amnesia. Overall, this study suggests TASIT is a valid measure for adolescents although it raised questions as to how effective normative data is for comparing performance in social cognition during childhood and adolescence. (JINS, 2013, 19, 1–11)

Social communication involves influencing what other people think and feel about themselves. We use the term conative theory of mind (ToM) to refer to communicative interactions involving one person trying to influence the mental and emotional state of another, paradigmatic examples of which are irony and empathy. This study reports how children with traumatic brain injury (TBI) understand ironic criticism and empathic praise, on a task requiring them to identify speaker belief and intention for direct conative speech acts involving literal truth, and indirect speech acts involving either ironic criticism or empathic praise. Participants were 71 children in the chronic state of a single TBI and 57 age- and gender-matched children with orthopedic injuries (OI). Group differences emerged on indirect speech acts involving conation (i.e., irony and empathy), but not on structurally and linguistically identical direct speech acts, suggesting specific deficits in this aspect of social cognition in school-age children with TBI. Deficits in children with mild-moderate TBI were less widespread and more selective than those of children with more severe injuries. Deficits in understanding the social, conative function of indirect speech acts like irony and empathy have widespread and deep implications for social function in children with TBI. (JINS, 2013, 19, 1–11)

The use of nonliteral language in clinical assessment, especially testing the patients’ ability to interpret proverbs, has a long tradition in psychiatry. However, its diagnostic sensitivity and specificity in dementias is not yet clear. The aim of this review article is to examine the current evidence on nonliteral/figurative language (proverb, metaphor, metonymy, idiom, irony, sarcasm) comprehension in Alzheimer's disease and related disorders. A comprehensive literature search identified 25 studies (16 proverb, 3 metaphor, 0 metonymy, 5 idiom, 3 sarcasm) on nonliteral language comprehension in dementia. Studies predominantly indicate a deficit. Most studies investigated Alzheimer's dementia. Applied correctly, nonliteral language is a worthwhile diagnostic tool to evaluate language and abstract thinking in dementias. During assessment, familiarity testing (e.g., by asking “are you familiar with the proverb XY”) is obligatory. Still, future research is needed in several areas: evidence on decline of nonliteral language over the course of the illness is limited. So far, almost no studies delineated proverb comprehension in high risk populations such as patients with mild cognitive impairment. Currently, there is a lack of studies addressing performance in direct comparison to relevant differential diagnosis like older-age depression, delirium, brain lesion, or other psychiatric conditions. (JINS, 2011, 17, 207–218)