Caption Translator

• A caption translator package based on open ai api
• demo at streamlit: https://miblue119-caption-translator-streamlit-app-7s6eqn.streamlit.app/

Usage

• install dependencies pip install -r requirements.txt

• Set your openai api key with export OPENAI_API_KEYOPENAI="sxxxxxxxx"

• Arguments

  • Necessary
    • --file_path: Set the transcript file path. Current only support .vtt and .srt format
  • Options
    • --text_engine: Set the open ai text engine. Default text engine is gpt-3.5-turbo
    • --language: Set the target translated language. Default language is japanese
      • support other language korean / german / traditional chinese / simplified chinese / french /dutch
        • please see the LANGUAGES definition at caption_translator/utils.py
    • --test: Whether to test part of content
    • --test_num: How many number of contents do you want to summarize?

• Example

  $python -m caption_translator.app --file_path ./examples/EP108_humanosis_Podcast.vtt --test --language ja
  

Resources

• https://blog.devgenius.io/creating-meeting-minutes-using-openai-gpt-3-api-f79e5fc15eb1
• https://blog.devgenius.io/counting-tokens-for-openai-gpt-3-api-59c8e0812eeb
• Open AI's open source tokenizer tiktoken: https://github.com/openai/tiktoken
  • Tokenization algorithm Byte Pair Encoding(1994 A New Algorithm for Data Compression) : https://zhuanlan.zhihu.com/p/424631681
    • data compression
  • Open AI cookbook's example to use tiktoken: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_count_tokens_with_tiktoken.ipynbhttps://huggingface.co/course/chapter6/5?fw=p
• Hugging face Tokenizer
  • BPE tokenization introduction: https://huggingface.co/course/chapter6/5?fw=pt
• Process long input
  • https://github.com/openai/openai-cookbook/blob/main/examples/Embedding_long_inputs.ipynb
  • https://github.com/openai/openai-cookbook/blob/main/examples/book_translation/translate_latex_book.ipynb
  • https://github.com/openai/openai-cookbook/blob/main/examples/fine-tuned_qa/olympics-1-collect-data.ipynb
  • oepnai web crawl qa: https://github.com/openai/openai-cookbook/blob/main/apps/web-crawl-q-and-a/web-qa.ipynb
  • splitting-chunking-large-input-text-for-summarisation-greater-than-4096-tokens
    • recursive summary: https://github.dev/daveshap/RecursiveSummarizer
    • other tools: https://github.com/miso-belica/sumy
• Others
  • https://help.openai.com/en/articles/5072263-how-do-i-use-stop-sequences

Prompt

• source: https://github.com/openai/openai-cookbook/blob/main/text_explanation_examples.md

Summarize the following text.

Text:
"""
Two independent experiments reported their results this morning at CERN, Europe's high-energy physics laboratory near Geneva in Switzerland. Both show convincing evidence of a new boson particle weighing around 125 gigaelectronvolts, which so far fits predictions of the Higgs previously made by theoretical physicists.

"As a layman I would say: 'I think we have it'. Would you agree?" Rolf-Dieter Heuer, CERN's director-general, asked the packed auditorium. The physicists assembled there burst into applause.
"""

Summary: