Sentiment Analysis: Detecting Valence, Emotions, and Other Affectual States from Text

The term sentiment analysis can be used to refer to many different, but related, problems. Most commonly, it is used to refer to the task of automatically determining the valence or polarity of a piece of text, whether it is positive, negative, or neutral. However, more generally, it refers to determining one’s attitude towards a particular target or topic. Here, attitude can mean an evaluative judgment, such as positive or negative, or an emotional or affectual attitude such as frustration, joy, anger, sadness, excitement, and so on. Note that some authors consider feelings to be the general category that includes attitude, emotions, moods, and other affectual states. Sentiment analysis can thus be considered as the task of automatically determining feelings from text. Russell (1980) developed a circumplex model of affect and showed that it can be characterized by two primary dimensions: valence (positive and negative dimension) and arousal (degree of reactivity to stimulus). Thus, it is not surprising that large amounts of work in sentiment analysis is focused on determining valence. However, there is some work on automatically detecting arousal, and growing interest in detecting emotions such as anger, frustration, sadness, and optimism in text. Further, the massive amounts of data emanating from social media have led to significant interest in detecting both valence and emotions in blog posts, tweets, instant messages, customer reviews, and Facebook posts. The vast majority of these approaches employ statistical machine learning techniques, although some rule-based approaches still persist. This chapter presents a survey of sentiment analysis research, covering statistical machine learning approaches for both valence and emotion detection from written text. (See surveys by El Ayadi, Kamel, and Karray (2011) and Anagnostopoulos, Iliou, and Giannoukos (2015) for an overview of emotion detection in speech. See Picard (2000) and Alm and Ovesdotter (2008) for a broader introduction of giving machines the ability to detect sentiment and emotions in various modalities such as text, speech, and vision.) We begin in Section 2 by discussing various challenges in sentiment analysis of text. In Section 3, we describe the diverse landscape of sentiment analysis problems, including detecting sentiment of the writer, reader, and other relevant entities; detecting sentiment from words, sentences, and documents; detecting stance towards events and entities which may or may not be explicitly mentioned in the text; detecting sentiment towards aspects of products; and detecting semantic roles of feelings. Many of the machine learning approaches for automatic detection of sentiment are supervised, that is, they first learn a model from a set of example instances labeled with