Xtag is an user interface based on Tcl/Tk/expectk languages. This interface integrates all the features of the tagger modules:

• the preparation
• the training
• the tagging
• the results.

The Help button allows to read the man pages with your favourite browser.
 

COMMAND:

>xtag  test.ref

In the main window you can select with the appropriate button the part of the tagger you want to execute.
You can follow the operations and have a look at the output.

CONFIG

But above all, you have to configurate the options of the text to process  in the "config" window with the usual  arguments of the tagger commands :

• Tag conversion List:  Specifies the tag conversion list. If this option is not specified then no tag substitutions will be applied.
• Word conversion list:  Specifies the word conversion list. If this option is not specified then no special tag substitutions are applied to individual words.
• Biases list: Specifies a set of transition biases to be be applied to the input matrices
• Fields Number: Number of fields preceding the [BOS|EOS] field (the default is 1).
• Lexical column:  Specifies the column where to find the word. It is assumed that the following field indicates if the word is sentence initial/final with "BOS|EOS" and the subsequent field contains the lexical annotations.
• Primary separator: Specifies the separator within [LEM,ANNOT] pair (the default is  '\'). rmk with the separator '\':  take in count that the backslash is processed twice, therefore type "\\\\".
• Secondary separator: Specifies the separator between the sets of ambiguous [LEM,ANNOT] pairs

assigned to a given word (default is '|').

PREPARATION
 

In the "Preparation" window, the text is prepare for the tagger or trainer programs but you can selecte the "Simple conversion" option that apply  the conversions without change the format (-H option).  See the mpreptxt man page.
 

TRAINING

In the "Training" window, you can make a training (mtrain program) but also operate  on the matrices (mcreate, mprint and edit). The matrices file can be created by mtrain which initializes the matrix with equi-proabable values based on the tags found in the corpus.The values can be readjusted to reflect user-defined preferences as stated in the biases_file. This training phase can be repeated for any number of iterations where each iteration may assign different probababilities. The matrices are used by the tagging program mtag to calculate the most probable tag for each word in a text.
mtrain readjusts the parameters and returns the new values in a compiled matrices file.
 

TAGGING
 

In the "Tagging" window you can make a tagging (mtag program) .The option "Re-estimate" allow to re-estimate the probalities to improve the accuracy. If the correct tag list and the matrices output file are indicated, the tagger will automatically readjust the values in the matrices according to the correct solutions and retag the text. This improve obviously the result for the given text, but can also improve the result for the text with the similar structures.
 

RESULTS
 

In the "Results" window, you can operate in the matrices with the biases rules (mbiases program), and print the results with the mdiff , mdiffb or  mcontext commands.
 

 
CONFIGURATION FILE
 

SEE ALSO

AUTHOR

Comments, suggestions, and bug reports are always welcome.