def process(self, ind, path):
tok = Tokenizer()
tok.tokenize_html(path)
self._titles.append(tok.title or u'Untitled')
for token, freq in tok._counter.items():
# Normalize by unique tokens.
self.globalht[token].append([ind, freq / len(tok._counter)])
def __init__(self, ds_name='nlvr2', split='train', task='ispeaker'):
self.ds_name = ds_name
self.split = split
self.data = json.load(
open(os.path.join(DATA_ROOT, self.ds_name, self.split + ".json"))
)
self.tok = Tokenizer()
self.tok.load(os.path.join(DATA_ROOT, self.ds_name, "vocab.txt"))
self.feat_mean = np.load(os.path.join(DATA_ROOT, self.ds_name, 'feat_mean.npy'))
self.feat_std = np.load(os.path.join(DATA_ROOT, self.ds_name, 'feat_std.npy'))
def top(self, query, num):
tok = Tokenizer()
tok.tokenize(query)
acc = Counter()
for token in tok._counter:
dict_entry = self.dict_entry(token)
print token, dict_entry
if not dict_entry:
continue
post_entries = self.post_entries(dict_entry[1], dict_entry[2])
# Adjust weights so that we don't open files multiple times in
# cases of repeated tokens. (e.g: dog cat dog cat dog)
for docid in post_entries:
post_entries[docid] *= tok._counter[token]
acc.update(post_entries)
return acc.most_common(num), len(acc)
class TestTokenizerGenAlphaDigit(
unittest.TestCase): # class Tokenizer, method tokenize_gen_alpha_digit
def setUp(self):
self.t = Tokenizer()
def test_empty_string(self):
s = list(self.t.tokenize_gen_alpha_digit(''))
self.assertEqual(len(s), 0)
self.assertEqual(s, [])
def test_no_spaces(self):
s = list(self.t.tokenize_gen_alpha_digit('мамамылараму'))
self.assertEqual(s, [Token('мамамылараму', 'alpha', 0, 12)])
def test_first_alpha(self):
s = list(self.t.tokenize_gen_alpha_digit('мама мыла'))
self.assertEqual(s[0], Token("мама", "alpha", 0, 4))
self.assertEqual(len(s), 4)
class TestTokenizerGen(unittest.TestCase):
def setUp(self):
self.t = Tokenizer()
def test_last_nonalpha(self):
s = list(self.t.tokenize_gen('мамамылараму2'))
self.assertEqual(len(s), 12)
self.assertEqual(s[11], Token('мамамылараму', 'alpha', 0, 13))
def test_first_alpha(self):
s = list(self.t.tokenize_gen('я иду в кино'))
self.assertEqual(s[0], Token("я", "alpha", 0, 1))
self.assertEqual(len(s), 7)
def test_empty_string(self):
s = list(self.t.tokenize_gen(''))
self.assertEqual(len(s), 0)
self.assertEqual(s, [])
def test_no_spaces(self):
s = list(self.t.tokenize_gen('яидувкиноcinema'))
self.assertEqual(s, [Token('яидувкиноcinema', 'alpha', 0, 15)])
self.assertEqual(len(s), 1)
def test_digital_string(self):
s = list(self.t.tokenize_gen('012345'))
self.assertEqual(len(s), 1)
self.assertEqual(s, [Token('012345', 'digit', 0, 6)])
def test_first_nonalpha(self):
s = list(self.t.tokenize_gen('!!!!я иду в кино cinema'))
self.assertEqual(len(s), 10)
self.assertEqual(s[0], Token('!!!!', 'punct', 0, 4))
def test_middle_nonapha(self):
s = list(self.t.tokenize_gen('я иду в кино00000 111 00000cinema'))
self.assertEqual(len(s), 13)
self.assertEqual(s[7], Token('00000', 'digit', 12, 17))
self.assertEqual(s[9], Token('111', 'digit', 18, 21))
import sys from tok import Tokenizer tok = Tokenizer() for review in sys.stdin: print ' '.join(tok.tokenize(review))
'nlvr2',
'spotdiff',
'adobe',
]
ds_root = "../dataset/"
for ds_name in DATASETS:
print("Processing dataset %s" % ds_name)
dataset = []
for split_name in ['train', 'valid']:
dataset.extend(
json.load(
open(os.path.join(ds_root, ds_name, split_name + ".json"))))
print("Finish Loading split %s" % split_name)
print("Number of data is %d." % len(dataset))
sents = sum(map(lambda x: x["sents"], dataset), [])
print("Number of sents is %d." % len(sents))
tok = Tokenizer()
tok.build_vocab(sents, min_occur=3)
tok.dump(os.path.join(ds_root, ds_name, "vocab.txt"))
wordXnum = list(tok.occur.items())
wordXnum = sorted(wordXnum, key=lambda x: x[1], reverse=True)
N = 50
print("Top %d Words:" % N)
for word, num in wordXnum[:N]:
print("%s: %d" % (word, num))
print()
op_code = (w >> 12) & 0o77
op_addr = w & 0xFFF
if w & (1 << 18) != 0: # address is extended
op_addr |= 0o70000
else: # long address command
op_code = ((w >> 15) & 0o37) + 48
op_addr = w & 0o77777
if op_indx == 0:
return f"{op_names[op_code]} {op_addr:>05o}"
else:
return f"{op_names[op_code]} {op_addr:>05o},M{op_indx:o}"
if __name__ == '__main__':
source = open(input_file).read()
t = Tokenizer(source)
PC = 0
DP = 0
irom = array.array('Q', [0xFFFFFFFFFFFFFFFF] * 65536)
dram = array.array('Q', [0xFFFFFFFFFFFFFFFF] * 32768)
fix_list = [] # tuples(pc, 'name', offset)
names = {}
bss_size = 0
error_count = 0
while not t.eof:
if keyword := t.get('IDENT'):
kwrd = keyword.val.lower()
line = keyword.line
import re
from collections import Counter
import just
import requests
from nearnlp.nearnlp import is_noun, is_verb, singularize
import functools
from tok import Tokenizer
from nltk.corpus import stopwords
from nostalgia.enrichers.google.custom_search import google_custom_search
ENGLISH_STOP = set(stopwords.words("english"))
t = Tokenizer(True)
t.drop("<b>", "remove html")
t.drop("<b/>", "remove html")
# can also use qoogle
interesting_keys = set()
for prefix in ["og:", "twitter:", ""]:
for key in [
"title",
"description",
"name",
"manufacturer_name",
"category_name_singular",
"long_description",
"snippet",
]:
interesting_keys.add(prefix + key)
def setUp(self):
self.t = Tokenizer()
def process(self, path):
tok = Tokenizer()
tok.tokenize_html(path)
self._counter.update(tok._counter)
self.write_tokens(path, tok._counter.keys())
