Bridging the Emotional Gap - From Objective Representations to Subjective Interpretations

Presented on 23 Oct at IJCCI and IC3K 2014

Keynote Lecturer: Marie-Jeanne Lesot

Abstract: In the framework of affective computing, emotion mining constitutes a classification task that aims at recognising the emotional content of various types of data including, but not limited to, texts, images or physiological signals. It adds to the traditional semantic gap, between low-level numerical data descriptions and their high-level conceptual interpretations, the difficulty of going from an objective to a subjective representation. After discussing the difficulty of a computational model of the labels to be considered in this specific classification task, due to the essential ambiguity and imprecision of emotions, the talk will illustrate the shift from numerical data representations to the emotions the data convey, through the integration of intermediate subjectivity levels, exploiting either external knowledge to include emotional information in the objective representation, or a subjective non-emotional level.