An IndoEuropeanSentenceModel is a heuristic sentence
designed primarily for English. Whether or not it balances
parentheses or forces the last token to be a boundary may be
set in the constructor. It uses the default implementation of
possible sentence starts and the following token sets:
any single letter
personal and professional titles, ranks, etc.
commas, colon, and quotes
times, months, etc.
U.S. political parties
U.S. states (not ME or IN)
possible stops (see above)
Note that all of these sets are case insensitive.
An Indo-European setence model is serializable. The model read
back in will be an instance of IndoEuropeanSentenceModel
with the same behavior as the model that was written.