Recently studied some of the Chinese affective analysis of some paper, very interested, and finally decided to write a Chinese affective analysis tool. As a loyal fan of open source thought, I also shortcoming a bad code to present to everyone, welcome everyone to make the decision brick. I hope this tool will bring us some practical use.
Currently, this tool only implements the use of a basic algorithm to predict the feelings of the article. After my test, basically satisfied with the Chinese sentence of emotional analysis, and accuracy can be trusted. If you have this demand, want to simply analyze the emotional tendencies of a set of Chinese comments, you can use this tool directly.
The following is a quantitative indicator of the performance and accuracy of this tool:
Performance: Processing about 100,000 Chinese characters per second
Accuracy: About 90%
I will maintain this tool for a long time, and will gradually add more prediction algorithms to increase the choice of more algorithms.
BitBucket Library Address: Https://bitbucket.org/shichaoqu/semantic-analysis-tool/overview
Features provided by the tool:
1. Based on Python-jieba Chinese word-wrapping, the article and sentence segmentation;
2. Use the Dalian University Emotion Analysis Thesaurus, to the article Word segmentation result carries on the emotion prediction;
3. Use Bsa_agorithm as the basic Affective Analysis algorithm, based on the emotional prediction of words to aggregate the entire article's emotional tendencies and emotional intensity.
TODO list:
1. Expand the emotional lexicon, the future will add hownet and Ntsu emotional word thesaurus support, and add a response to the emotional Word locator interface;
2. Extend affective analysis algorithm, support more common algorithm choice, provide more accurate emotion analysis algorithm;
3. Emotional information extraction, extract the views of holders, subject and emotional statements, and their relationship.