Question and answer task matching model based on quantum measurement and self-attention mechanism

The invention discloses a question and answer task matching model based on quantum measurement and a self-attention mechanism. One-hot vectors are used for representing different semantic units, words are composed of a group of mutually orthogonal semantic units, sef-attention is introduced for modeling semantic weights among different words, sentences are modeled by a density matrix, a hybrid system containing a plurality of words is represented, and the two mixing systems are projected to the same plane through the measurement matrix so as to judge the similarity degree of the two sentences. Compared with a reference model, the model verifies the effectiveness in a question and answer task. Compared with an original quantum language model, each part of the model has good physical meaning and mathematical constraint, and experimental analysis shows that the model has higher interpretability. The question and answer matching effect is improved. 本发明公开一种基于量子测量与自注意力机制的问答任务匹配模型，使用one-hot向量表示不同的语义单元，单词由一组相互正交的语义单元