This issue was created by revision r878.
Previously, some Meta Collectors only recorded an ngram once per document.
Unsystematically, other Meta Collectors recorded the ngram's count per
document. This makes a difference because most users will constrain their list
of ngram features based on ngram frequency, and the choice of method changes
which ngrams are most frequent.
Now, Meta Collectors are standardized to record the count of an ngram in a text.
Note: This bugfix may change your experiment results.
Original issue reported on code.google.com by EmilyKJa...@gmail.com on 9 Jun 2014 at 10:39
Original issue reported on code.google.com by
EmilyKJa...@gmail.com
on 9 Jun 2014 at 10:39