Knowledge retrieval

Finding the relevant fragment is the beginning of understanding.

A fragment and its surroundings

Knowledge retrieval brings relevant material into the moment when it might help. In Sato, the conceptual system represents textbook chunks as 768-dimensional embeddings stored locally in ChromaDB.

A retrieved chunk is not the whole context. The question is how to carry enough of its source and surroundings along with it, and how to connect it to the learner’s own questions.

A working entry. The connections may change.

↳ /connected to

Sato

knowledge retrieval → forms a layer of

Where knowledge comes from

knowledge retrieval → needs to preserve

The machine-readable web

knowledge retrieval → depends on legibility in