ddi_kernels.py 文件源码

def get_svm_train_line(tree, pair, sid):
    lmtzr = WordNetLemmatizer()
    e1id = compact_id(pair.eids[0])
    e2id = compact_id(pair.eids[1])
    tree = tree.replace(pair.entities[0].tokens[0].text, 'candidatedrug')
    tree = tree.replace(pair.entities[1].tokens[0].text, 'candidatedrug')
    #tree = tree.replace(sid.replace('.', '').replace('-', '') + 'e', 'otherdrug')
    sid2 = compact_id(sid) + 'e'
    # TODO: replace other entities
    #tree = rext.sub(sid2 + r'\d+', 'otherdrug', tree)
    #print "tree2:", tree
    if tree[0] != '(':
        tree = '(S (' + tree + ' NN))'
    #this depends on the version of nlkt
    ptree = Tree.fromstring(tree)
    #ptree = Tree.parse(tree)
    leaves = list(ptree.pos())
    lemmaleaves = []
    for t in leaves:
        pos = get_wordnet_pos(t[1])
        lemma = lmtzr.lemmatize(t[0].lower(), pos)
        lemmaleaves.append(lemma)
    #lemmaleaves = [ for t in leaves)]
    logging.debug("tree:" + tree)
    line = '1 '
    line += '|BT|'  + tree
    #bowline = '(BOW (' + ' *)('.join(lemmaleaves) + ' *)) '
    #ptree = Tree.parse(bowline)
    #ptree = ptree.pprint(indent=-1000)
    #bowline = ptree.replace('\n', ' ')
    #bowline = '|BT| ' + bowline
    #if not bowline.count("otherdrug") > 8:
    #    line += bowline
    #else:
        #print "problem with BOW!"
    #line += bowline
    line += '|ET| '

    #i = 1
    #for m in docsp[ddi.PAIR_SSM_VECTOR]:
    #    line += " %s:%s" % (i, m)
    #    i += 1
    #line += " 2:" + str()
    #line += " |EV|"
    line += '\n'
    return line