Aphasia Communication Cards

Using vector search technology to find words when you can't remember them

Try these example searches:

Or with typos (still works!):

Search History

No search history yet. Try searching for something!

All Cards

No matching cards found. Try different descriptive words.

How This Works

This demo simulates a vector search system for aphasia patients:

  1. When a card is added to the system, an on-device LLM generates a semantic vector embedding capturing its meaning.
  2. These embeddings are stored in a vector database (e.g., SQLite with vector extension).
  3. When a user types a search query, the LLM generates a vector for that query.
  4. The system finds cards with embeddings most similar to the query vector.
  5. This semantic matching works even with misspellings or when users can't remember the exact word.

Note: This demo uses simulated vectors - a real implementation would generate these using the on-device LLM.