Pos Tagger

• Train a tagger:

  • java -jar SemiSupervisedTagger.jar train -input [input-file] -model [model-file]

    • Other Options:

        * -dev [dev-file]  dev file address
        * -cluster [cluster-file]  brown cluster file address
        * -viterbi   if you want to use Viterbi decoding (default: beam decoding)
        * -update:[mode]  for beam training; three #modes: max_viol, early, standard (default: max_viol)
        * -delim [delim]   put delimiter string in [delim] for word tag separator (default _) e.g. -delim /
        * beam:[#b]  put a number [#b] for beam size (default:5); e.g. beam:10
        * iter:[#i]  put a number [#i] for training iterations (default:20); e.g. iter:10
      
        * NOTE: in every iteration the model file for that iteration will have the format [model-file].iter_#iter e.g. model.iter_3
      

• Tag a file:

  • java -jar SemiSupervisedTagger.jar tag -input [input-file] -model [model-file] -output [output-file]
    • Other Options:
    • -delim [delim] put delimiter string in [delim] for word tag separator (default _) e.g. -delim /

• Tag a partially tagged file:

  • java -jar SemiSupervisedTagger.jar partial_tag -input [input-file] -model [model-file] -output [output-file]
    • For words with no tag information, put *** as the tag; e.g. After_IN our_*** discussion_*** ._.
    • Other Options:
      • -delim [delim] put delimiter string in [delim] for word tag separator (default _) e.g. -delim /