INNER CODE UNIT · Python
main
RQLuo/MixTeX-Latex-OCR · mixtex_data_gen/gen.py:162
def main(input_text_file, input_tex_file, output_folder):
# Step 1: Clean the input text file by removing non-English characters
cleaned_text_file = 'en_only.txt'
remove_non_english_characters(input_text_file, cleaned_text_file)
# Step 2: Extract LaTeX formulas from the input LaTeX file
formulas = extract_latex_formulas(input_tex_file)
# Step 3: Process the cleaned text and insert LaTeX formulas randomly
processed_text_file = 'en_line.txt'
process_text(cleaned_text_file, processed_text_file, formulas)
# Step 4: Read in the processed text and further format it with LaTeX commands
with open(processed_text_file, 'r', encoding='utf-8') as f:
txt_content = f.read()
with open(cleaned_text_file, 'r', encoding='utf-8') as f:
lines = f.readlines()
txt_content = remove_symbols(txt_content)