Graph Search is a natural language search engine introduced by Facebook in 2013 that aims to answer user queries with relevant information rather than just returning links. It uses a graph database model where entities like users, pages, posts are represented as nodes and their connections as edges. This allows it to understand relationships between entities and power semantic searches. The architecture includes components for entity recognition, semantic parsing, and ranking results based on their relevance to the query using algorithms like N-shortest path and EdgeRank.