Why is Weka GUI output different from Java code?

Why is that the result from running the filter StringToWordVector in Weka GUI is different from the equivalent java code? I use the same attributes as I used in the gui but the tokenizer in java doesn't seem to do a proper job! I was told by a Ph.D student that it is common and no further answer from him.

Please help. My project is stalled.

Here is my code:

DataSource tempSource = new DataSource("/home/r_omio/Dataset.arff");
Instances temp = tempSource.getDataSet();
NumericToBinary nbTemp = new NumericToBinary();
nbTemp.setInputFormat(temp);
temp = Filter.useFilter(temp, nbTemp);
StringToWordVector stringFilterTemp = new StringToWordVector(2500);

stringFilterTemp.setOptions( 
   开发者_JAVA技巧 weka.core.Utils.splitOptions("-R 1,2,3,4 -W 2500 -prune-rate -1.0 <br>-N 1 -stemmer weka.core.stemmers.NullStemmer -M 1 -tokenizer weka.core.tokenizers.WordTokenizer -delimiters \" \\r\\n\\t.,;:\\\'\\\"()?![]_\"")
 );


stringFilterTemp.setInputFormat(temp);
temp = Filter.useFilter(temp, stringFilterTemp);

I suspect your delimiters are incorrectly escaped. Try using the default delimiters in the GUI and leaving the tokenizer out in Java, which will use the default, and see if you get the same value.