Sentiment analysis also referred to as opinion mining, is an automated process for identifying and classifying subjective information such as sentiments from a piece of text usually comments and reviews. Supported by machine learning algorithms, it is possible to identify positive, neutral or negative opinions, being possible to rank or classify them in order to reach some kind of conclusion or obtain any type of information. Thus, this paper aims to perform a systematic literature review in order to report the state-of-the-art of machine learning techniques for sentiment analysis applied to texts of reviews, comments and evaluations of scientific papers.