This work details appraisal extraction from attitude expressions. Here, by attitude expressions, we refer to those single words that convey the evaluation of sentiments or emotional states, about human behaviors, objects, processes or people, according to the Appraisal Theory of language. The attitude words can be classified into affect, judgment, and appreciation; either positive or negative. Extraction of the attitude words has a significant range of applications from opinion extraction and summarization, up to temporal opinion analysis. To determine the attitude, we use two machine learning techniques; namely, Support Vector Machines and Random Forest. These algorithms classify a given word starting from a vector that represents the information from the context where the words tend to occur. On the other hand, we can observe the context of the words relying on a corpus of sentences from user generated contents, such as reviews, editorials and other online texts.