ChatGPT for Translation | ChatGPT用于翻译
This is a simple tool that uses ChatGPT to translate text into a specified target language in a faithful way to the original. The tool takes a text file (.pdf, .txt, .md, .html and .rtf) or a folder of text files as input, and outputs a translated text file or a bilingual text file with the original and translated text side by side. Special optimization has been done especially for academic paper PDF parsing and translation.
使用ChatGPT将文本以忠于原文的方式翻译成指定的目标语言。该工具接受一个文本文件（.pdf, .txt, .md, .html或.rtf）或者一个包含文本的文件夹，并生成一个直接翻译后的文本或一个双语的(并列显示原始文本和翻译文本)文本。尤其对于学术论文 PDF 解析和翻译做了特别的优化。
Use this on Google Colab (recommended). See here
Google Colab上使用这个工具(推荐)。见这里
Note: Use absolute paths instead of relative paths to process PDFs.
注意：处理 PDF 的时候使用绝对路径而不是相对路径。
Simple Example | 简单例子
git clone https://github.com/Raychanan/ChatGPT-for-Translation.git
cd ./ChatGPT-for-Translation/
pip install -r requirements.txt --quiet
python ChatGPT-translate.py --input_path=input.txt --openai_key=password
This command will translate the text in input.txt into simplified Chinese using ChatGPT. You can also specify any language you want. For example, --target_language="Japanese". See this txt as an example.
这个命令将使用ChatGPT把input.txt中的文本翻译成简体中文。你也可以指定任何你想要的语言。例如，--target_language="Japanese"。翻译后的txt文件例子见这里
# Azure's OpenAI service, use this:
python ChatGPT-translate.py --input_path=input.txt --use_azure --azure_endpoint=endpoint_uri --azure_deployment_name=deployment_name --openai_key=your_AOAI_key
Translate Folder Files | 翻译文件夹内所有的文本文件
python ChatGPT-translate.py --input_path=./folder/ --openai_key=password
Bilingual Translation Example | 双语翻译例子
python ChatGPT-translate.py --bilingual --input_path=input.txt --openai_key=password
Prerequisites | 要求
You need a OpenAI API key (https://beta.openai.com/signup/)
你需要一个OpenAI的API密钥（https://beta.openai.com/signup/）
Arguments | 可用参数
--num_threads: The number of threads to use for translation (default: 5).
--only_process_this_file_extension For example, set only_process_this_file_extension="txt"
--not_to_translate_people_names Whether or not to translate names in the text. This can be useful if you are translating academic texts. By default, names will be translated.
--not_to_translate_references By default, not to translate references.
--keep_first_two_paragraphs Keep the first three paragraphs of the original text. By default, false.
Acknowledge
PDF parser is based on scipdf project on Github. Some adjustments were done to allow users to parse PDFs without having to initialzing a server locally.