""" only replace head nouns """
from functools import reduce
from os.path import join as pjoin
from pattern import *
import spacy
from word2number import w2n
import argparse
import os
# load spacy
nlp = spacy.load('en')
def detect_number(text):
left = 0
detect_results = list()
while left < len(text):
right = left + 1
if text[left].pos_ == 'NUM':
right = left + 1
while right < len(text) and text[right].pos_ == 'NUM':
right += 1
if right == len(text):
continue
number = ' '.join([word.text for word in text[left: right]])
if not valid(number):
left = right
continue
num_range = [left, right]
while right < len(text) and text[right].pos_ in ['ADJ', 'CCONJ']:
right += 1
if right == len(text):
continue
if text[right].pos_ == 'NOUN':
while right < len(text) and text[right].pos_ == 'NOUN':
right += 1
n = text[right - 1].text
n_index = right - 1
else:
left = right
continue
detect_results.append([number, num_range, n, n_index])
elif text[left].pos_ == 'DET' and (text[left].text == 'an' or text[left].text == 'a'):
right = left + 1
num_range = [left, right]
while right < len(text) and text[right].pos_ in ['ADJ', 'CCONJ']:
right += 1
if right == len(text):
continue
if text[right].pos_ == 'NOUN':
while right < len(text) and text[right].pos_ == 'NOUN':
right += 1
n = text[right - 1].text
n_index = right - 1
else:
left = right
continue
detect_results.append([text[left].text, num_range, n, n_index])
left = right
return detect_results
def valid(word):
# check number
try:
numbers = [w2n.word_to_num(it) for it in word.split()]
if len(numbers) > 1:
return False
except ValueError:
return False
# check English
chars = map(lambda c: 'a' <= c <= 'z', word)
return reduce(lambda a, b: a or b, chars, False)
prefix_list = ['test', 'dev', 'train']
# rule:
# 1. NUM string
# 2. after NUM should be NOUN
# 3. Noun described by NUM should be the last one of the first NOUN string after NUM
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--data_name', type=str, default='resnet152_numex_precomp')
parser.add_argument('--data_path', type=str, default='../data/')
args = parser.parse_args()
# load all coco captions
coco_captions = list()
number_set = set()
for prefix in prefix_list:
os.system('mkdir {:s}/{:s}/{:s}_ex'.format(args.data_path, args.data_name, prefix))
path_coco_captions = pjoin('{:s}/{:s}/'.format(args.data_path, args.data_name), prefix + '_caps.txt')
coco_captions += open(path_coco_captions).readlines()
# count word-frequency for nouns and adjectives
frequency = dict()
number_set = {'two', 'six', 'eight', 'five', 'fifteen',
'nine', 'a', 'an', 'four', 'three',
'seven', 'fourteen', 'twenty', 'eighteen', 'ten',
'one', 'thirty', 'eleven', 'twelve', 'thirteen',
'sixteen', 'seventeen', 'nineteen', 'sixty', 'forty',
'fifty', 'seventy', 'eighty'}
# process_numbers
number_dict = dict()
for item in number_set:
if item in ['a', 'an', 'one']:
number_dict[item] = 1
else:
try:
number_dict[item] = w2n.word_to_num(item)
except ValueError:
print('value error:', item)
# replace these words with words which are similar to themselves but have different meanings
for prefix in prefix_list:
path_coco_captions = pjoin('{:s}/{:s}/'.format(args.data_path, args.data_name), prefix + '_caps.txt')
real_captions = open(path_coco_captions).readlines()
for i, caption in enumerate(real_captions):
replacement_cnt = 0
fout = open('{:s}/{:s}/{:s}_ex/{:d}.txt'.format(args.data_path, args.data_name, prefix, i), 'w')
text = nlp(caption.lower())
number_info = detect_number(text)
# replace
for item in number_info:
number_range = item[1]
w = item[0]
noun = item[2]
noun_index = item[3]
if w not in number_dict:
continue
# enumerate all possible words for replacement
for r in number_dict:
if number_dict[r] == number_dict[w]:
continue
word_list = [x.text for x in text]
# singularize or pluralize
if number_dict[r] == 1:
new_noun = singularize(noun)
elif number_dict[w] == 1:
new_noun = pluralize(noun)
else:
new_noun = noun
word_list[noun_index] = new_noun
new_word_list = word_list[:number_range[0]] + [r] + word_list[number_range[1]:]
replacement = ' '.join(new_word_list)
fout.write(replacement)
replacement_cnt += 1
if prefix in ['train', 'dev']:
for j in range(replacement_cnt, 5):
fout.write(' '.join(['<unk>' for _ in caption.strip().split()]) + '\n')
fout.close()
print(prefix, i, replacement_cnt)
