Retrieval-Augmented Generation
Definition
Retrieval-Augmented Generation (RAG): A technical framework that combines external knowledge retrieval with large language model generation to improve answer accuracy and timeliness.
Usage Example
An enterprise RAG system enables large models to generate accurate responses based on up-to-date business knowledge by retrieving from internal document repositories.
Related Terms
Contribution Details
Related terminology
Attention Mechanism
A technique that enables a model to dynamically attend to different parts of an input sequence by computing correlations between queries and keys/values to perform information weighting.
Word Embedding
Word embeddings map words to low-dimensional continuous vector spaces, placing semantically similar words closer together. Word2Vec and GloVe are classic embedding methods that laid the foundation for subsequent language models.
Named Entity Recognition
Named Entity Recognition (NER) is a foundational task in natural language processing that identifies meaningful entities—such as people, locations, organizations, and time expressions—from unstructured text. It is a key technology for information extraction and knowledge graph construction.