Storing And Retrieving Data Using Vector Embeddings

The invention describes a system for storing and finding digital information more effectively. When new data is saved, a specialized computer program, specifically a transformer neural network, analyzes it to create a unique numerical fingerprint called a vector embedding. This fingerprint is then stored alongside the original data. When someone searches for information, the system uses these numerical fingerprints to quickly find relevant data, rather than just matching keywords. The claims specifically narrow this process to using a transformer neural network for creating these fingerprints.