andreasvc

maxlen 15 unfolded False arity marks True binarized collinize right h=1 v=1 tailmarker markovize rank > 3 estimator dop1
python -u runexp.py  56312.42s user 296.32s system 96% cpu 16:20:51.28 total
../disco-dop/interp0.dop
labeled f-measure   : 76.12293144208039
unlabeled f-measure : 79.19621749408984
../disco-dop/interp1.dop
labeled f-measure   : 68.59122401847574
unlabeled f-measure : 72.74826789838338
../disco-dop/interp2.dop
labeled f-measure   : 69.07723459647092

andreasvc

Mary is really very happy --> is Mary really very happy?
0.50 S --> NP VP        [(0, 1)]
0.50 S --> VP_2 NP      [(0, 1, 0)]
1.00 VP --> V VP|<ADV>  [(0, 1)]
1.00 VP_2 --> V VP|<ADV>	[(0,), (1,)]
0.50 VP|<ADV> --> ADV VP|<ADV>  [(0, 1)]
0.50 VP|<ADV> --> ADV ADJ       [(0, 1)]
0.50 ADJ --> Epsilon    ['happy']
0.50 ADJ --> Epsilon    ['sad']
0.50 ADV --> Epsilon    ['really']

andreasvc

import time
import numpy as np
cimport numpy as np
from libc.stdlib cimport malloc, free
from cpython.array cimport array, clone

cdef long N = 1000000
cdef double* ptr
cdef array ar, template = array('d')

andreasvc

The D 1 D@1-1 1 D@0-1 1
dog N 0.5   N@0-2 1
cat N@1-2 1 N 0.5
barks   V 0.5   V@0-4 1
meows   V 0.5   V@1-4 1
loudly  RB 1    RB@0-5 1    RB@1-5 1

andreasvc

import sys, collections
def rotations(a):
    return {a[x:] + a [:x] for x in range(1, len(a))}

lexicon = collections.defaultdict(set)
for a in open(sys.argv[1]):
    lexicon[len(a) - 1].add(a.strip())
for length in sorted(lexicon):
    if length == 1:
        continue

andreasvc

andreasvc

cdef extern from "macros.h":
    # test whether the b'th bit of array a is set:
    unsigned long TESTBIT(unsigned long a[], int b)

cdef unsigned long foo[2]
foo[0] = 281474976710656UL
foo[1] = 0

# what the macro does:
print foo[0] & (1UL << 48)

andreasvc

def cartpi(seq):
  """ A depth-first cartesian product for a sequence of iterables;
  i.e., all values of the last iterable are consumed before advancing the
  preceding ones. Like itertools.product(), but supports infinite sequences.

  >>> from itertools import islice, count
  >>> list(islice(cartpi([count(), count()]), 9))
  [(0, 0), (0, 1), (0, 2), (0, 3), (0, 4), (0, 5), (0, 6), (0, 7), (0, 8)]
  """
  if seq:

andreasvc

""" Match lines in one file with those of another,
and produce line numbers. """
import io
import sys

USAGE = """Match lines in one file with those of another, and get line numbers.
usage: python %s sents text output
where sents and text are files with one sentence per line.
The result will be of the form "1|line", written to file "output".
Everything is assumed to be encoded with UTF-8.""" % sys.argv[0]

andreasvc

""" Compare different strategies for adding a large number of small log
probabilities. """
from __future__ import print_function
from math import log, exp, fsum, isinf
from random import expovariate

N = 10000


def logprobadd(x, y):