typst_hlm / exinfo.py
fasdfsa's picture
不再仔细校对文字和标点符号,给以后解决
08daf36
Raw
History Blame Contribute Delete
5.22 kB
# 改成原地加换行空格标点符号
if __name__ == '__main__':
import glob
import os
import json
import numpy as np
from pathlib import Path
import opencc
s2t = opencc.OpenCC('s2t.json') # 简繁转换
t2s = opencc.OpenCC('t2s.json')
chs = glob.glob('data/ch*', recursive=False)
m5s = {}
with open('data/variants.json', "r", encoding='UTF-8') as fp:
variants = json.load(fp)
for i, pth in enumerate(chs):
pth = pth.replace('\\', '/')
parts = pth.split('/')
chapter = parts[1]
out2 = f'data/{chapter}/out2'
with open(f"{out2}/page_texts.json", "r", encoding='UTF-8') as fp:
page_texts = json.load(fp)
with open(f"{out2}/pageIds_pageNames.json", "r", encoding='UTF-8') as fp:
pageIds_pageNames = json.load(fp)
pageNames_pageIds = {}
for k, v in pageIds_pageNames.items():
pageNames_pageIds[v] = k
boxs = glob.glob(f"{out2}/*.boxs", recursive=False)
last_x = -1
last_y = -1
last_height = -1
new_line = False
new_space = False
lines = []
line = []
signs = ['。', ':', ',', '?', '、', '!']
for j, pt in enumerate(boxs):
sign_count = 0
stem = Path(pt).stem
pageId = pageNames_pageIds[stem]
page_text = page_texts[pageId]
page_text = page_text.replace(' ', '')
pt_boxsnew = pt.replace('.boxs', '.boxsnew')
if os.path.exists(pt_boxsnew):
continue
with open(pt, "r", encoding='UTF-8') as fp:
box = json.load(fp)
g = 0
for k, bx in enumerate(box):
charPoly = bx["charPoly"]
charText = bx["charText"]
bx["sign"] = ''
bx["space"] = ''
bx["is_menu"] = False
bx["is_menutitle"] = False
sign = ''
if charText in variants:
charText = variants[charText]
charText = s2t.convert(charText)
g = k + sign_count
if g < len(page_text) and page_text[g] in signs:
sign = page_text[g]
# 把符号加到前一个字
box[k-1]["sign"] = sign
sign_count += 1
g = k + sign_count
if g < len(page_text) and page_text[g] in variants: # 异体
before = page_text[:g]
after = page_text[g+1:]
oldchar = page_text[g]
newchar = variants[oldchar]
page_text = before + newchar + after
if g < len(page_text):
before = page_text[:g]
after = page_text[g+1:]
oldchar = page_text[g]
newchar = s2t.convert(oldchar) # 简繁
page_text = before + newchar + after
if page_text[g] == charText:
pass
else:
pass
lu = [charPoly["x0"], charPoly["y0"]]
ru = [charPoly["x1"], charPoly["y1"]]
rd = [charPoly["x2"], charPoly["y2"]]
ld = [charPoly["x3"], charPoly["y3"]]
width = max(ru[0], rd[0]) - min(lu[0], ld[0])
height = max( ld[1], rd[1] ) - min(lu[1], ru[1])
points = np.array([lu, ru, rd, ld])
if last_x == -1:
last_x = lu[0]
else:
x_diff = abs(lu[0] - last_x)
if x_diff > int(width / 2):
new_line = True
if new_line:
# 把换行符加到前一个字
box[k-1]["sign"] = '\n'
lines.append(line)
line = []
new_line = False
if last_y == -1:
last_y = lu[1]
last_height = height
else:
y_diff = abs(lu[1] - last_y)
if y_diff > last_height + last_height / 2:
new_space = True
if new_space:
# 把空格加到前一个字'
new_space = False
line.append(bx)
last_x = lu[0]
last_y = lu[1]
last_height = height
box_text = ''
for k, bx in enumerate(box):
charPoly = bx["charPoly"]
charText = bx["charText"]
sign = ''
if 'sign' in bx:
sign = bx["sign"]
box_text = box_text + charText + sign
with open(pt_boxsnew, 'w', encoding='utf-8') as fp:
json.dump(box, fp, indent=4, ensure_ascii=False)
print(f"box {j+1} / {len(boxs)}")
print(f"{chapter} {i+1} / {len(chs)}")