Coursera Flash Sale
40% Off Coursera Plus for 3 Months!
Grab it
Explore the classification of textual documents according to emotions through both supervised and unsupervised approaches in this 57-minute conference talk. Learn about the challenges of emotion annotation in text, particularly how emotions can be subtle, context-dependent, and vary across domains and cultures, with special attention to low-resourced languages like Croatian. Discover a novel methodology for automatic extraction of emotion lexicons using s-discordance measures from symbolic data analysis and its application to semi-supervised classification in vector space models. Examine potential applications of symbolic data analysis for intrinsic evaluation of word representations in neural language models and for emotion-based document classification using modern language models. Gain insights into psychological models that underpin emotion analysis and understand the complexities of working with annotated datasets versus emotion lexicons in natural language processing tasks.
