The packets used for classification have some extraneous words:
- "10" and "points" appears in all tossups
- Words in pronunciation guides also appear, but these wouldn't be too useful for classification, since pronunciation guides differ per packets
Consider putting "10" and "points" as stop words. Filtering out pronunciation guides is trickier. You may need a parameter that takes in what character is used to denote a pronunciation guide (braces or square brackets).