This paper evaluates the performance of classical machine learning algorithms—Multinomial Naive Bayes, Support Vector Machines, and Artificial Neural Networks—in classifying excerpts from Philippine Supreme Court decisions based on their legal topics. Such classification is vital for generating comprehensive and organized case syllabi, which are crucial for legal education and research. For this study, we constructed a labeled dataset composed of jurisprudence excerpts with corresponding labels extracted from syllabi provided by the Philippine Reports in their publications of Supreme Court cases. To address the data imbalance encountered, we applied the SMOTE. For feature extraction, each labeled excerpt underwent TF-IDF processing. Dimensionality reduction using SVD was applied to the inputs of both the SVM and ANN models. The resulting vectors, representing the excerpts, were then used to train the SVM, MNB, and ANN models. Finally, we used PCA to visualize the excerpts. Preliminary results indicate that while both SVM and ANN achieve satisfactory performance with similar accuracy, MNB outperforms both, exhibiting the best overall performance among the said classifiers. While there is room for further improvement, the current MNB model is already useful for the syllabus tagging process.