Xiaopeng Motors announces the patent for a large-scale acoustics and semantics model, which can improve the model's response speed.

Tianyancha asset clue information shows that recently, the patent application for "acoustic semantic large model, server, voice interaction method and storage medium" filed by Guangzhou Xiaopeng Automotive Technology Co., Ltd. has been published. The abstract reveals that this application discloses an acoustic semantic large model, server, voice interaction method, and computer-readable storage medium. The acoustic semantic large model includes acoustic encoding module, character transcription module, knowledge retrieval module, and large language model module. The acoustic encoding module is configured to generate an acoustic feature vector of the voice request based on the input voice request. The character transcription module is configured to transcribe the voice request into a corresponding character sequence, where the character sequence includes characters corresponding to the words in the voice request. The knowledge retrieval module is configured to retrieve supplementary information from an external knowledge base based on the character sequence. The large language model module is configured to determine the natural language processing results based on the acoustic feature vector and the supplementary information. Thus, through an end-to-end acoustic semantic large model, the serial processing of multiple modules is reduced, the processing delay of voice requests is decreased, the model response speed is improved, and user experience is enhanced.