In this talk we will describe the attempts made at making machines "more human" by giving them sentiment and politeness abilities. We will give a perspective on automatic sentiment and emotion analysis, with a description of our work in this area, touching upon the challenging problems of sarcasm arising from numbers, and multitask and multimodal sentiment and emotion analysis. Subsequently we touch upon the interesting problem of "making computers polite" and our recent work on this. We will end with noting the places of rule based, classical ML based and deep learning-based approaches in NLP.
展开▼
