Chinese Sentiment Analysis base on dictionary and rules.
prior to v0.0.4, bixin depends on cppjieba-py
, which requires a c++ 11 compillation makes hard to use, I decided to use jieba_fast
.
it will solve the following problems:
- hard to install the dependency
cppjieba-py
- can't load user dictionary
- word segment difference from
jieba
but it slower than use cppjieba-py
> pip3 install bixin
from bixin import predict
text ="幸福每时每刻都会像路边的乞丐一样出现在你面前。要是你觉得你所梦想的幸福不是这样的,因而断言你的幸福已死亡,你只接受符合你的原则和心愿的幸福,那么你就会落得不幸。"
# 出自安德烈·纪德《人间食粮》
predict(text)
# sentiment score: 0.42
sentiment score is in the range of -1 to 1
predict
will load dictionary data at first time,to load it manually use predict.classifier.initialize()
Test with 6226 taged corpus mixed up with shopping reviews 、Sina Weibo tweets 、hotel reviews 、news and financial news
accuracy: 0.827771
Notice:neutral texts are all ignored.
details about test dataset see wiki 关于测试数据集
> pip3 install -e ".[dev]" git+https://github.com/bung87/bixin
./dictionaries dictionaries from vary sources
./data processed dictionaries through ./scripts/tagger.py
./scripts/release_data.py release data to package
./scripts/score.py
all data archives: https://github.com/bung87/bixin/releases/tag/v0.0.1
run accuray testing with all .txt files under test_data directory sentence per line end with a space and a tag n or p
nosetests -c nose.cfg
for single python version
tox
for multiple python versions
bixin was inspired by dongyuanxin's DictEmotionAlgorithm
MIT © bung