Multimedia fragment search system and method based on large language model

The invention relates to the technical field of artificial intelligence, in particular to a multimedia fragment search system and method based on a large language model, and the system is composed of a data collection module, a model fine tuning module, a user interaction module, a voice recognition module, a semantic matching module and a user interaction module. The multimedia fragment searching system and method based on the large language model have the advantages that semantic matching is conducted on spoken language data by means of the powerful semantic understanding capacity of the large language model, the method is more accurate and efficient, and the defect that the spoken language data cannot be accurately matched through a traditional method is overcome. In addition, according to the method, the user can directly listen to or check the audio clips related to the search content of the user, and a more visual and comprehensive search result is provided. According to the method, accurate and efficie