unisense

a homemade python package for sentiment analysis

installation

python setup.py install

requirement

• scikit-learn

usage

from unisense.en import EnLogRegSentiment

a toy model is provided in ./unisense/test/en/model/

# create sentiment class
test = EnLogRegSentiment(model_path='./toy_model.pkl')

return the sentiment score, input should be list of string

# text analysis
test.analysis(['Hello world!'])

for each document, the score is tuple of neg and pos probability

example

rule based sentiment model for english

from unisense.en import EnRuleSentiment
model = EnRuleSentiment()
vocab = {'hello': -1, 'word': -1, 
         'hello world': -1, 'world hello': -1}
model.train(vocab)
model.create_matcher()
model.kw_match(['hello World word'])

exception

• model not exist
• vocab format error
• input format error

test

pytest --pyargs unisense

doc

• en
  • EnLogRegSentiment
  • EnRuleSentiment
• fr
  • FrRuleSentiment

Generally, the sentiment class has the following functions

• train()
• load_model()
• save_model()
• analysis()

to be done