Do not update newly created model properties
There is no real reason to favor old behavior over the current, plus
the new behavior saves an extra disk write (which are expensive).
Patch by: Sverre Rabbelier
# Performance note: I benchmarked this code using a set instead of
# a list for the stopwords and was surprised to find that the list
# performed /better/ than the set - maybe because it's only a small
# list.
stopwords = '''
i
a
an
are
as
at
be
by
for
from
how
in
is
it
of
on
or
that
the
this
to
was
what
when
where
'''.split()
def strip_stopwords(sentence):
"Removes stopwords - also normalizes whitespace"
words = sentence.split()
sentence = []
for word in words:
if word.lower() not in stopwords:
sentence.append(word)
return u' '.join(sentence)