HyDE is an approach for information retrieval and Retrieval augumented generation (RAG), where search is enhanced by using an artificially created answer the actual query by the user. Usually, an LLM generates this synthetic answer, which is then embedded and queried against the embedding space of the knowledge. Thus not only the question query can be used for retrieval, but also an answer embedding.

https://docs.haystack.deepset.ai/docs/hypothetical-document-embeddings-hyde