Proslang
AI slang support for real conversations. A smartwatch + phone concept that helps non-native speakers capture unfamiliar slang in the moment — and understand it later.
Course project · team
Knowing English doesn't always mean following the conversation.
Slang, idioms, and tone often depend on context. In live conversations, the moment to ask or search can pass quickly.
From phone, to watch, to both.
Pivot 1 · Phone → watch. Pulling out a phone mid-conversation feels slow and impolite. So we moved everything to the watch.
Pivot 2 · Watch → watch + phone. In the first round of testing with 5 international students, a watch doing everything was hard to use and hard to read. So we split the job: quick support on the watch, deeper learning on the phone.
Quick meaning and one-tap save.
Examples, context, origin, and review.
See it catch slang mid-conversation.
How Proslang helps users capture unfamiliar slang on the watch, confirm the meaning in the moment, and review context later on the phone.
From live conversation to deeper understanding.
Capture on Watch
Recognize unfamiliar slang or idioms during live conversations directly from the watch.
Voice detection · Non-intrusive
Confirm in the Moment
Show a quick meaning so users can decide whether to save, edit, or ignore the phrase.
Quick meaning · One-tap save
Review on Phone
Saved phrases sync to the phone for examples, cultural context, origin, and usage patterns.
Examples · Context · Origin
Every decision made it lighter and clearer.
Testing showed a crowded watch was hard to read. Now it only shows the meaning and a one-tap save.
Improves readability and makes the learning experience feel lighter.
Because the product uses voice recognition in social settings, privacy became a key concern. A stronger version should make listening status visible, ask for explicit permission, and let users control when recognition starts and stops.
This was a 2022 project. Today, AI could make it actually understand context.
A real prototype, tested in real chats
Connect live speech-to-text and a language model, then test it in real conversations instead of a clickable mock-up.
Meaning for this moment
A language model can read the whole sentence, so "slaying" is explained for this conversation, not as a dictionary entry.
Listen only when asked
Process speech on the device where possible, and make listening always visible and easy to stop.