Multi-person long speech semantic recognition and abstract generation method, system and device and medium

The invention provides a multi-person long voice semantic recognition and abstract generation method, system and device and a medium, and the method comprises the steps: carrying out the voice signal enhancement of a noisy voice signal through a Deducs model, and obtaining a real voice signal; voiceprint features are extracted from the real voice signals through an ERes2Net model, a personal voiceprint library is formed based on the voiceprint features, and identities of different speakers are recognized; inputting a real voice signal into the trained Conformer end-to-end model for voice content recognition, matching the identity of a speaker, and converting the speaking content into a text; carrying out abstract generation on the speaking content text through a large language model; according to the method, the context information of the long text can be effectively captured, the attention degree of local information is high, and the technical problem that technical terms or contexts in the specific field ar