Week 1: Foundations
Day 2: Why Sign Language AI Is Different
Goal
Understand why sign-language AI is harder than text-only AI.
Learn
- ASL is visual, spatial, and grammatical. It is not English words shown with the hands.
- Meaning can depend on handshape, palm orientation, movement path, location, facial expression, body shift, eye gaze, timing, and the surrounding conversation.
- A text-only model mostly sees tokens. A sign-language model has to work with motion over time, space around the body, and multiple channels happening at once.
Example
- A yes/no question in ASL may use facial grammar such as raised eyebrows and a forward head position. If a system tracks only hand motion, it may miss that the sentence is a question.
- A sign near the forehead can mean something different from a similar hand motion near the chest. Location matters.
Practice
- Choose one familiar sign or phrase.
- List what the camera must capture: hands, face, body, timing, space, and lighting.
- Circle anything that would be hard for a computer to detect if the signer moves quickly or leaves the frame.
Checkpoint
Before moving on
You can explain why ASL cannot be treated as English on the hands.
Quality note
Quality note
A dataset with clear hands but missing facial information may still be weak for ASL because grammar can live in the face and body.