Two-Tower Model
A neural network architecture with two independent encoder pathways that process different data types into a shared vector space for fast comparison.
Think of It Like This
Like an interpreter who translates English into a universal sign language and French into the exact same sign language so both sides can communicate.
Also known as a bi-encoder, one 'tower' encodes a user query while the other entirely separate 'tower' encodes the documents or items. Because the towers are disconnected, document embeddings can be precomputed and cached offline. At inference time, the user query is quickly encoded and compared via dot product for rapid retrieval.