Unsafe sentence tokenizer in sentiment analysis

[renoust]

Hi!

In jSentiments.py, in polarScores_text(), you are processing each sentence by:

for sent in text.split(u'。'):
   etc.

This part actually crashes when you have an empty sentence coming in, that we can protect using:

for sent in text.split(u'。'):
    if len(sent.strip()) == 0:
         continue
    etc.