Embedding-based Retrieval in Facebook Search
The unified embedding framework developed to model semantic embeddings for personalized search, and the system to serve embedding-based retrieval in a typical search system based on an inverted index are introduced.
Abstract
Search in social networks such as Facebook poses different challenges than in\nclassical web search: besides the query text, it is important to take into\naccount the searcher's context to provide relevant results. Their social graph\nis an integral part of this context and is a unique aspect of Facebook search.\nWhile embedding-based retrieval (EBR) has been applied in eb search engines for\nyears, Facebook search was still mainly based on a Boolean matching model. In\nthis paper, we discuss the techniques for applying EBR to a Facebook Search\nsystem. We introduce the unified embedding framework developed to model\nsemantic embeddings for personalized search, and the system to serve\nembedding-based retrieval in a typical search system based on an inverted\nindex. We discuss various tricks and experiences on end-to-end optimization of\nthe whole system, including ANN parameter tuning and full-stack optimization.\nFinally, we present our progress on two selected advanced topics about\nmodeling. We evaluated EBR on verticals for Facebook Search with significant\nmetrics gains observed in online A/B experiments. We believe this paper will\nprovide useful insights and experiences to help people on developing\nembedding-based retrieval systems in search engines.\n