Sentic neural networks: A novel cognitive model for affective common sense reasoning

In human cognition, the capacity to reason and make decisions is strictly dependent on our common sense knowledge about the world and our inner emotional states: we call this ability affective common sense reasoning. In previous works, graph mining and multi-dimensionality reduction techniques have been employed in attempt to emulate such a process and, hence, to semantically and affectively analyze natural language text. In this work, we exploit a novel cognitive model based on the combined use of principal component analysis and artificial neural networks to perform reasoning on a knowledge base obtained by merging a graph representation of common sense with a linguistic resource for the lexical representation of affect. Results show a noticeable improvement in emotion recognition from natural language text and pave the way for more bio-inspired approaches to the emulation of affective common sense reasoning.