Automatic Question Classifiers: A Look at Oracle SQL Exam Questions

Imagine you're preparing for an Oracle SQL certification exam. The questions you face might contain code snippets, making them tricky to classify. This study takes a closer look at designing automatic classifiers for such questions. Why is this important? Classifying questions helps in creating a relevant test paper from a large test bank. But when questions involve code snippets, it gets tougher.

The study used factorial experiments to figure out what makes a good classifier. They tested different ways to represent features and various machine learning methods. Interestingly, using a scheme called TF-IDF along with Logistic Regression proved best for overall performance. But for precision, going with Support Vector Machines was better. The key takeaway? The way features are represented matters most, followed by the chosen machine learning method.

Actions