Google Introduces Speech-to-Retrieval (S2R) Approach that Maps a Spoken Query Directly to an Embedding and Retrieves Information without First Converting Speech to Text