Learning to Filter Spam E-Mail: A Comparison of a Naive Bayesian and a Memory-Based Approach