Setting up Method 2.

Note: Basic knowledge of GATE is assumed (loading and working with corpora and different processing resources).

1) Linguistic preprocessing:
For this method the lingusitic preprocessing has to be done outside GATE.

First, run your corpus  thorugh the Minipar dependency parser.
To get the  best results you should preprocess your corpus as follows:
          * transform all  UpperCase to LowerCase.
          * if there is a verb in the first position and ends in "s" - remove this termination.
          * split up the text so that you have one sentence per line.

Second, transform the Minipar output in annotations readable by GATE.  A java class that does this is available here. Minipar's output format is slightly different on different platforms so you might need to re-adapt this file.

2) Pattern based extraction:
Then load the following JAPE transducers one by one in this order and add them to the pipeline.
Alternatively you can simply load main.jape which will load all the transducers for you.
          * 1_myWordTransfer (set inputASName =  Original markups and outputASName = Tokens)
          * 2_mySentenceTransfer (set inputASName =  Original markups and outputASName = Tokens)
          * 2_DoubleObjectCorrection  (set inputASName = outputASName = Tokens)
          * 3_NPIdentifyer (set inputASName = outputASName = Tokens)
          * 4_ConjunctiveNouns (set inputASName = outputASName = Tokens)
          * 5_ConjunctiveVerbs (set inputASName = outputASName = Tokens)
          * 6_VerbObjectIdentifyer (set inputASName = outputASName = Tokens)

Note: you could save this corpus somewhere so that you do not repeat these steps.




3) Ontology learning - this step is the same as for Method1