2019/05/08 by Soujanya Poria, Poria, Soujanya, Navonil Majumder +5 · 17 citations
Computer Science · Psychology · #Affective computing #Artificial Intelligence (cs.AI) #Artificial intelligence #Communication #Computation and Language (cs.CL) #Computer science #Conversation #Data science #Emotion and Mood Recognition #Emotion recognition #FOS: Computer and information sciences #Field (mathematics) #Key (lock) #Mental Health via Writing #Psychology #Scalability #Sentiment Analysis and Opinion Mining #cs.AI #cs.CL
paper · pdf · doi:10.48550/arxiv.1905.02947
published in arXiv (Cornell University) (Cornell University)
arxiv created 2019/05/08 · openalex publication_date 2019/05/08 · arxiv updated 2019/05/09 · openalex created_date 2022/07/29 · openalex updated_date 2026/08/06
Emotion is intrinsic to humans and consequently emotion understanding is a\nkey part of human-like artificial intelligence (AI). Emotion recognition in\nconversation (ERC) is becoming increasingly popular as a new research frontier\nin natural language processing (NLP) due to its ability to mine opinions from\nthe plethora of publicly available conversational data in platforms such as\nFacebook, Youtube, Reddit, Twitter, and others. Moreover, it has potential\napplications in health-care systems (as a tool for psychological analysis),\neducation (understanding student frustration) and more. Additionally, ERC is\nalso extremely important for generating emotion-aware dialogues that require an\nunderstanding of the user's emotions. Catering to these needs calls for\neffective and scalable conversational emotion-recognition algorithms. However,\nit is a strenuous problem to solve because of several research challenges. In\nthis paper, we discuss these challenges and shed light on the recent research\nin this field. We also describe the drawbacks of these approaches and discuss\nthe reasons why they fail to successfully overcome the research challenges in\nERC.\n