Finding Surprisingly Frequent Patterns Of Variable Lengths In Sequence Data