The aim of this study is to investigate how the emotional orientations in peer assessment affect student performance. Given that the emotional data in peer comments exist in the form of unstructured text, this study applied the text sentiment analysis method to the mining and analysis of peer-review texts. For this purpose, emotional orientations are quantified as emotional variables with a polarity, including positive, negative, and neutral, and degrees of intensity, including high, moderate, and mild. Specifically, a semester-long course entailing repeated peer assessments was organized, and eight hundred and eighty-five valid comments were collected. By using quantitative analysis of the qualitative emotional variables, this paper further explores the relationships between emotional orientations and learning performance. The research results showed that emotional orientations had statistically significant effects on student performance: students with a negative emotional orientation and high emotional intensity were more likely to obtain better grades, while students with average grades preferred positive evaluations and presented no obvious changes in emotional intensity. Accordingly, several recommendations are summarized for researchers and instructors.