The teaching and learning process in current lectures can be done only by attending online classes through the Massive Open Online Course (MOOC). But in practice, learners find it difficult to find an appropriate course since its subject is not complemented with adequate descriptions. When uploading a new course material, the instructors tend to be reluctant to clearly define the course's descriptions, learning outcomes, and course matter. They would be likely only to upload a set of sentences that cover these things. This paper explains the method of extracting learning content using classification then automatically adds annotations to the learning content. The annotation label contains a course name, description, learning outcomes, and course matters. The dataset was obtained from a set of learning contents in Bahasa Indonesia. It was classified using four methods, rule-based implementation without machine learning, Machine Learning (ML) implementation with Random Forest, Support Vector Machine, and Naive Bayes. The non-ML classification method produced the worst result with an accuracy value of 71.7%. However, the best result was obtained from the ML with Random Forest Classifier. We implemented this method to train the over-sampled training data and hit an accuracy value of 93.3%. Besides, the model was able to produce appropriate annotation output from the new testing data.