To elaborate a bit: the task was to look for frequent patterns in headlines of news articles, in hopes of identifying repetitive daily (hourly, weekly) reports that would tend to have formulaic content -- i.e. to isolate groups of headlines like:

TOP STORIES FOR WEDNESDAY, JUNE 14
TOP STORIES FOR TUESDAY, JULY 27
...
STOCK INDEXES, NYSE 11:00 GMT
STOCK INDEXES, NASDAQ 13:00 GMT
...
BASEBALL: FINAL SCORES ...
...
[download]

Those are made-up examples, but you get the idea: find bunches of headlines in a multi-year archive that start with the same words. Of course, the same code (with a different split regex) would create the more typical character-based trie structure.