README.md

RITE_zh-CN

Recognizing Inference in Text of Simplified Chinese

Steps

1. preprocessing with xml and format normalization
2. processing on grammar: segment, pos tagging, named entity recognize
3. extract multifeatures with knowledge
4. send to classifier and train or test
5. post-processing: rule regular

And the modules in code:

Usage

• cd ./Source
  python run.py

• You can edit run.py to train or test any file

  • Just test the file 'test.txt' and output accuracy and Macro F

    if __name__ == '__main__':
        test(prefix='test')

  • test a pair of text that input with user

    if __name__ == '__main__':
        test_show()

  • test the file 'rite_test_new_without_label.txt' and save the predict label

    if __name__ == '__main__':
        test_raw_pair(prefix='rite_test_new_without_label')

  • train with the file 'train.txt'

    if __name__ == '__main__':
        train(gamma=0.4, prefix='train')

    ​