import nltk
from nltk.corpus import conll2000

grammar = r'''
    NP: {<DT>?<POS>?<CD>?<PRP$>?<JJ[RS]*>*<VBG>*<RB[R]*><NN>}
        {<DT>?<POS>?<CD>?<PRP$>?<JJ[RS]*>*<VBG>*<NN[PS]*><CC><DT>?<POS>?<CD>?<PRP$>?<JJ[RS]*>*<VBG>*<NN[PS]*>+} # coordination
        {<DT>?<POS>?<CD>?<JJ[RS]*>*<VBG>*<NN[PS]*>+} # compounds
        {<PRP>} # pronoun
        {<WDT>} # that
        {<EX>} # it/there expletive
'''

cp = nltk.RegexpParser(grammar)

training_chunks = conll2000.chunked_sents('train.txt', chunk_types=['NP'])
test_chunks = conll2000.chunked_sents('test.txt', chunk_types=['NP'])

print('Evaluering på treningskorpuset:\n', cp.evaluate(training_chunks))
print('\nEvaluering på testkorpuset:\n', cp.evaluate(test_chunks))