September 2022

Himansh Sheoran received his MS Dual Degree in  Computer Science and Engineering (CSE). His research work was supervised by Dr. Priyanka Srivastav. Here’s a summary of his research work on Encoding of emotional concepts in depression:

Depression has always been an exciting topic of research due to its high prevalence and complex underlying mechanisms. Cognitive impairment, alterations in mood, emotion dysregulation, aversion to activity are just a few of the consequences of depression. Despite there being an ocean of literature present there, attempting to understand depression, we have just scratched the surface. There is still a lot to explore about the intricacies of this perplexing subject. This thesis is yet another attempt to understand and evaluate how depression changes how we understand and represent information (or concepts). In this work, our primary focus is on emotional concepts. In a nutshell, we are trying to look for answers to the question, ’Whether or not there exist any differences in the way people exhibiting moderate to severe depressive symptoms encode emotional concepts, compared to people with no or minimal depressive symptoms?’. To address this question, we first conducted a pilot study involving 48 participants. A corpus of 50 emotional words of varying hedonic and arousing content was given to the participants in this study. They were instructed to rate their affective experience of valence and arousal for each word on a 9-point self-assessment manikin (SAM) scale. This study aimed to record the Indian affective experience of emotional words and construct a comprehensive affective word database for the following main study. Fifteen words were selected from the corpus, which according to the participants’ responses, were capable of accurately capturing the required affect and arousal. These 15 words were clustered on the arousal-valence bi-dimensional model into five groups (3 belonging to each) and used as stimuli for the second study. An additional objective of the study was to compare and contrast the affective responses of the Indian population with the data collected by other studies on the western people to see if there existed any cultural differences between them. Following the pilot study, we conducted a detailed psychological study involving 107 participants and instructed them to generate properties for 15 emotional words (from the pilot study) presented to them as stimuli. Using state-of-the-art computational models, we evaluated the data collected to efficaciously shed some light on the linguistic, semantic, and sentiment differences in the properties generated by the two groups of samples (participants exhibiting no or mild symptoms and participants showing moderate to severe symptoms). These two groups were created after assessing the predisposition emotional state of the entire sample individually by using Beck Depression Inventory-II (BDI). We found an overall increased negative sentiment and decreased positive sentiment in the properties generated by the sample showing depressive symptoms. Emotion analysis of the data reflected that the sample with depressive symptoms displayed higher levels of anger and sadness and lower levels of joy and optimism. Analyzing the semantics of the properties showed that the same sample used fewer social, motivational, and affective words. Moreover, an independent linguistic probing of the data showed that higher symptoms resulted in increased usage of first-person singular pronouns (i.e., ’I’) and negation words, and, lower use of adjectives. Thus, through the aforementioned extensive analysis, this research attempts to point to the fact that there does seem to be a difference in the way emotional information is encoded in people exhibiting depressive symptoms. The current result will strengthen the conceptual understanding of identifying risks to depression in young adults and help design better health care services on college campuses.