For the course "Information Retrieval and Extraction", Fall semester, 2017. Writen by Hao-Yung Chan (Student ID: R04521618)

Python 3.6 (3.6.3 is used in development)

  $ git clone https://github.com/katrina376/ir2017-prj2.git
  $ cd ir2017-prj2

  $ python3 feature-base.py <PATH_TO_TRAIN_DATA> <PATH_TO_TEST_DATA> <PATH_TO_SEG_CORPUS> <WIDTH_OF_NOUN_WINDOW> <WIDTH_OF_VERB_WINDOW>

Be aware that <PATH_TO_SEG_CORPUS> should be the segmented version of the corpus.

<WIDTH_OF_NOUN_WINDOW> and <WIDTH_OF_VERB_WINDOW> are the parameter of the model. In the experiment, the odds in the interval of 3–13 are used.

Make sure that meta.py is in the same folder as feature-base.py. meta.py contains the glossary of POS tags to their simplified version. (Source:中研院平衡語料庫詞類標記集)

For example, train.txt, test.txt, Dream_of_the_Red_Chamber_seg.txt are all in the same folder as feature-base.py, set the width of noun window to 5 and the width of verb window to 7. Run:

  $ python3 feature-base.py train.txt test.txt Dream_of_the_Red_Chamber_seg.txt 5 7