Anlamsal Arama
Search method that brings results based on query instead of keyword match.
Meaning search (semantic search) is a search method that results not only by passing the same words among documents, but by making a match based on understanding underlying expressions. Traditional search engines can also find a document written with different words, such as meaningal search "automobile repair", while only finding documents with these full words, with the question "how to repair the car". This is possible to convert queries and documents to high-dimensional numerical vectors through a embedding model; meaning close texts in this vector space are also close to geometrically.
The meaningal search is one of the basic components of today’s artificial intelligence applications, especially the Retrieval-Augmented Generation (RAG) systems: when asked a large language model, the model lists and binding parts related to a wide collection of documents with meaningal search before producing the answer, so the model can also produce accurate answers based on non-in-education data or non-updated information. The vector databases such as Pinecone, Weaviate and FAISS are specially designed infrastructures to make quick sense search between millions of embedding and have a wide range of applications from modern search engines to recommendation systems.
The meaningal search is one of the basic components of today’s artificial intelligence applications, especially the Retrieval-Augmented Generation (RAG) systems: when asked a large language model, the model lists and binding parts related to a wide collection of documents with meaningal search before producing the answer, so the model can also produce accurate answers based on non-in-education data or non-updated information. The vector databases such as Pinecone, Weaviate and FAISS are specially designed infrastructures to make quick sense search between millions of embedding and have a wide range of applications from modern search engines to recommendation systems.