README.md

metadata

title: Arabic Sign Language NLP API
emoji: 🤟
colorFrom: green
colorTo: blue
sdk: docker
app_port: 7860
pinned: false

Arabic Sign Language NLP API

Translates Arabic text (Fus-ha and Ammiya) into sign animation sequences.

Endpoints

Method	Path	Description
GET	/	Health check — returns model info and sign count
POST	/translate	Translate Arabic text (JSON body)
GET	/translate?text=...	Quick translate via URL param
GET	/signs	List all signs in the database
GET	/sequence-file	Read the last saved sequence file

POST /translate

{
  "text": "انا عايز اروح المدرسة",
  "save_sequence": false
}

Response:

{
  "status": "success",
  "input_text": "انا عايز اروح المدرسة",
  "sequence": ["انا", "يريد", "يذهب", "مدرسة"],
  "total_steps": 4,
  "sign_count": 4,
  "letter_count": 0,
  "missing_keypoint_files": [],
  "detailed_plan": [...]
}

Setup

1. Upload your arabic_sign_lang_features.csv to the Space files.
2. (Optional) Upload your keypoints/ folder for .npy validation.
3. Set CSV_PATH env variable if your CSV has a different name.

Environment Variables

Variable	Default	Description
CSV_PATH	arabic_sign_lang_features.csv	Path to sign label CSV
KEYPOINTS_FOLDER	keypoints	Folder with .npy files
SIMILARITY_THRESHOLD	0.72	AraBERT match threshold

Interactive Docs

Visit /docs for the Swagger UI.