Multi-dimensional sentiment classification in online learning environment

Text-based sentiment analysis as a tool for monitoring online learning environment has elicited increasing interesting and been widely used in practice. Correctly identifying author sentiment in a stream of text presents a number of challenges including accurate language parsing, differing perspectives between author and reader, and the general difficulty in accurately classifying natural language semantics. This paper documents the development and initial results of a unique multi-dimensional sentiment analysis agent for online learning environment, in order to provide overall student feedback on a number of different levels as well as identify potential problems during the delivery of the course. This sentiment analysis agent monitors student interaction in the messaging, discussion and collaboration tools found in the Moodle learning environment, and classifies textual data into one of six dimensions: positive, negative, neutral, insightful, angry, and joke. Ultimately we see this work being especially useful to larger digital learning environments - especially massive open online courses (MOOCs) - where instructors and administrators are unable to read every individual forum or discussion item, but require a way to identify significant changes in tone and sentiment in order to quickly address potential students or user issues.