Embedding
Definition
A dense vector representation of text, audio, or other data that captures semantic meaning in a continuous space.
Embeddings map discrete inputs (words, sentences, audio clips) to continuous vectors in a high-dimensional space where semantically similar items are positioned near each other. Word embeddings like Word2Vec and GloVe represent individual words; sentence embeddings from models like BERT represent entire passages.
In speech recognition, audio embeddings represent acoustic features in a learned space. Speaker embeddings capture voice characteristics for diarization. In text processing, embeddings enable semantic search, clustering, and similarity comparison. The embedding space learned by large models captures nuanced relationships between concepts.