Method and system for realizing dialogue emotion recognition based on dynamically extracting sub-dialogues

The invention provides a dialogue emotion recognition method and system based on dynamic extraction of sub-dialogues, and the method comprises the steps: capturing the context feature expression of a dialogue and the feature expression of a speaker through a bidirectional long-short-term memory network, and obtaining the dialogue feature expression of each speaker; in the dialogue, the sub-dialogue range of each utterance in the dialogue is selected according to the relative position of the historical utterance and the target utterance of each speaker in the dialogue, in addition, the same sentence always expresses different emotions in different theme scenes, and a plurality of different themes exist in a section of dialogue in an energy-saving manner; the theme of each sub-dialogue is extracted by adopting variational self-coding to assist in identifying the emotion of the target utterance; the attention score of each utterance in the dialogue is obtained by adopting a multi-head attention mechanism to enha