yanshuyuan / classify Goto Github PK

1. Ready for train set and test set:
   The 'data' directory contains many subject subdirectory, such as 'ent', 'game' and so on,
   every subdirectory contains a 'link.dat' file and many shell script and python script.
   
   'link.dat' record the url of many html texts.
	
   so, if you want to classify for a subject, you should enter the 'subject' directory, 

   and run: './auto.sh'

   the 'auto.sh' will help you download the html text and parse words, 

   and save the a file into directory 'train' and 'test'.

2. Clean unused train set and test set file
   The test set and test set file may have some rabbish text, 

   if you want delete them, you could do like this:
	
   run: './clean.sh'

   the 'clean.sh' will delete the file which bytes is less then 300bytes.

3. Create train data file by bayes:

   The 'train.py' is a python script for creating train data,
  
   you should run: 'python train.py'

   the python script will read the train set file from directory 'train',
 
   and output the train data file 'data.dat' .

4. Test

   The 'test.py' is a test script for test the classify effect,

   you should run: 'python test.py'

   the python script will read the test set file from directory 'test',

   and output the effect.

5. Classify for a url link
   
   The 'classify.sh' is a shell script which could classify the url link into a subject of train data file,

   you should run: 'classify.sh url'

   the script will tell you, it may be belong to which subject.