ABSTRACT

In this chapter we provide an overview of sentiment analysis, a discipline that has evolved over the last decade as a branch of natural language processing, at the core of what is known as “opinion mining”. It has attracted great attention, both in academic research and commercial applications, due to the demand for computational tools that offer reliable metrics on users’ opinion of products and services, which they express daily on a multitude of online sites and 512social media platforms. Sentiment analysis seeks to automatically obtain an input text’s qualification on a scale of a certain precision (binary or otherwise) according to its axiology, that is, the degree of positivity or negativity it expresses. We introduce the fundamental concepts in sentiment analysis where textual analysis, and particularly corpus linguistics methods and techniques, play a fundamental role. We define the various approaches according to the aim they pursue: sentence-level, document-level or aspect-based classification, and according to the types of algorithm employed, both by lexicon-based and machine learning approaches, which may use supervised and unsupervised techniques. Finally, we explore in detail the crucial role that the corpus has in all of these approaches and methodologies, both in supervised ones, for the extraction of classifying features, and in unsupervised methods, such as those based on distributional semantics (word embeddings).