import nltk, re, pprint

def read(filename):
    """Takes file formatted like sentences, returns list of pairs, one for each sentence."""
    raw = open(filename).read()
    lines = raw.split('\n')[::-1]
    data = []
    while lines:
        line = lines.pop().split('#')[0].strip() # strip comments

        if not line: continue

        words = line.lower().replace('.', '').split()
        tags = lines.pop().split('#')[0].strip().split()
        glosses = lines.pop().split('#')[0].strip().split()
        trans = lines.pop().split('#')[0].strip()

        new_item = zip(words, tags, glosses), trans

        data.append(new_item)

        continue

    return data

def sents(data):
    """Takes list like that returned by read(), returns list of lists of word/tag pairs."""
    return [[word[:2] for word in sentence[0]] for sentence in data]

data = read('sentences')
sentences = sents(data)

grammar = r"""
  NP: {<NN>?<JJ>*<DT>?}           # Chunk sequences of NN, JJ, DT
  NPR: {<NP>?<SRel>}              # Chunk NPs with relative clauses
  SRel: {<VBR><NP.*>?<NP.*>?}     # Chunk sequences of VBR and 0-2 NPs
  """
cp = nltk.RegexpParser(grammar, loop=2)

for s in sentences: print cp.parse(s)