Emotion recognition has emerged as an important research area which may reveal some valuable input to a variety of purposes. People express their emotions directly or indirectly through their speech, facial expressions, gestures or writings. Many different sources of information, such as speech, text and visual can be used to analyze emotions. Nowadays, writings take many forms of social media posts, micro-blogs, news articles, etc., and the content of these posts can be useful resource for text mining to discover and unhide various aspects, including emotions. Extracting emotions behind these postings is an immense and complicated task. To tackle this problem, researchers from diverse fields are trying to find an efficient way to more precisely detect human emotions from various sources, including text and speech. In this sense, different word-based and sentence-based techniques, machine learning, natural language processing methods, etc., have been used to achieve better accuracy. Analyzing emotions can be helpful in many different domains. One such domain is human computer interaction. With the help of emotion recognition, computers can make better decisions to help users. With the increase in popularity of robotic research, emotion recognition will also help making human–robot interaction more natural. This survey covers existing emotion detection research efforts, emotion models, emotion datasets, emotion detection techniques, their features, limitations and some possible future directions. We focus on reviewing research efforts analyzing emotions based on text and speech. We investigated different feature sets that have been used in existing methodologies. We summarize basic achievements in the field and highlight possible extensions for better outcome.
