metadata

title: EngTexToASLGloss
emoji: 📊
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 3.40.1
app_file: eng_to_aslGloss_app.py
pinned: false

How to use English to ASL Gloss

Introduction

This is a program that uses GPT4 to translate English text to ASL gloss. This program makes use of prompt engineering to first describe the task it wants GPT4 to complete (along with examples of translations of English text and it's ASL gloss translations), and then providing GPT4 with the English text to translate to ASL_gloss.

These prompts are included as part of the package.

Requirements

This program makes use of the following packages:

Python 10 environment (required for tiktoken and openAI)
openAI
tiktoken
openAI key and access to GPT4

Updates and improvements

As GPT4 improves, we will continue to update our model.
The most recent update (2/27/2024) uses of the most current version of GPT4-turbo (gpt-4-0125-preview). This model includes an expanded token output as well as improved insturction handling and reduction of 'laziness'. - Version 2.0 (GPT4-Turbo, 2/27/2024) - Improved ASL gloss translation - Directional verbs (e.g. HELP-you) - Fingerspelling (Seems to know better what to fingerspell)

Limitations

Token length
- Althoughg GPT4 has a token limit, GPT4-Turbo has increased the number of tokens it can process and return, which improves our text translation length possibilities. As a result, this previous limitation is grately improved. That said, if your text is still longer than the token limit, this program will not translate your text. (The program will calculate tokens for you, so you don't need to worry about doing that conversion)
  - As this limit continues to improve, we will update the program accordingly.
Speed and Cost
- Be aware that each time we run this code, it re-prompts GPT4 (since we can't save our models). As a result, it takes a little time and it costs a little bit of money. But it's not too bad.
Imperfect ASL gloss
- Although this does a pretty good job, it's not perfect ASL gloss
- Some things you should watch for include:
  - word order
  - fingerspelling
  - formatting rules <>
- As we continue to improve the prompting, we will update with any new prompts that we create that improve the ASL gloss output.

Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference