Quantum measurement-based computer medium for question and answer task matching

Abstract

The invention discloses a question and answer task matching method based on quantum measurement. A measurement matrix encoder and a question and answer matching module are included. The language encoder is used for encoding a current question and answer sentence to obtain a sentence encoding matrix; the structure comprises an embedded layer and a middle layer, the embedding layer is used for initializing word codes; the middle layer is used for further calculating word codes to obtain a density matrix of words so as to further obtain a density matrix of question and answer sentences; the measurement matrix encoder is mainly used for encoding observation information of a current sentence and obtaining a measurement encoding matrix of a question and answer sentence; the question and answer matching module is used for judging whether the currently input question sentence is matched with the answer sentence; based on the end-to-end quantum language model, the problem that interaction information cannot be considered in the quantum language model is solved, and the question and answer matching effect is improved.

Description

technical field [0001] The invention belongs to the computer field of question-answer matching tasks for data mining, and in particular relates to a computer medium for question-answer task matching based on quantum measurement. Background technique [0002] Language model is a very important part of natural language processing. Language models can be divided into classical probabilistic language models and quantum probabilistic language models. Quantum probabilistic language models rely on quantum probability and quantum phenomena. Findings suggest quantum phenomena in human cognition [1] , especially in language comprehension [2] . Uncertainty in quantum probability can be used to better represent uncertainty in sentence semantics [3] . According to Van Rijsbergen, quantum theory can be understood through mathematical formalisms in logical, geometric and probabilistic information retrieval models [4] . It inspired the later combination of quantum theory and natural ...

AI Technical Summary

Problems solved by technology

[0004] However, NNQLM still does not take into account the importance of the interaction between the question and the answer

If there is no interaction, the matching effect will not be strong

Images