Sentiment analysis in social media is harder than in other types of text due to limitations such as abbreviations, jargon, and references to existing content or concepts. Nevertheless, social media provides more information beyond text, such as linked media, user reactions, and relations between users. We refer to this information as social context. Recent works have successfully leveraged the fusion of text with social context for sentiment analysis tasks. However, these works are usually limited to specific aspects of social context, and there have not been any attempts to analyze and apply social context systematically. This work aims to bridge this gap by providing three main contributions: 1) a formal definition of social context; 2) a framework for classifying and comparing approaches that use social context; 3) a review of existing works based on the defined framework.