Intelligent question answering method and device for multi-hop reasoning question

Abstract

The embodiment of the invention discloses an intelligent question answering method and device for a multi-hop reasoning question. The method comprises the steps of obtaining a question text; performing semantic coding on the question text to obtain semantic coding representation of the question text; according to the semantic coding representation of the question text, determining a first prediction result, wherein the first prediction result is a prediction result of the position where at least one question body of the question text is located; determining a second prediction result according to the semantic coding representation of the question text, wherein the second prediction result is a prediction result of at least one question relationship of the question text; according to the first prediction result and the second prediction result, generating a sub-question text, wherein the sub-question text comprises at least one sub-question; according to a screening document, answering at least one sub-question in sequence, and obtaining an answer corresponding to the at least one sub-question, wherein the screening document comprises the answer corresponding to the at least one sub-question; and determining a final answer of the question text according to the answer corresponding to the at least one sub-question.

Description

technical field [0001] The invention relates to the technical field of natural language processing. In particular, it relates to an intelligent question answering method and device for multi-hop reasoning questions. Background technique [0002] The multi-hop reasoning system is an important technology for intelligently analyzing complex question descriptions and finding multiple clues for answers in intelligent question-and-answer business scenarios. It is an important part of intelligent question-and-answer business scenarios. [0003] Deep learning technology has made great progress in traditional single-hop reading comprehension tasks, achieving a level comparable to that of humans, but many single-hop reading comprehension tasks can be answered by simply matching sentences with paragraphs of questions, and do not involve complicated reasoning. [0004] In order to let the machine learn the real reasoning ability, many high-quality multi-hop question answering datasets...

AI Technical Summary

Problems solved by technology

Although these works perform well on many tasks, they also have limitations that cannot be ignored

[0006] First of all, the internal reasoning mechanism of the end-to-end reading comprehension model is not clear. They usually use an additional discriminator to judge whether a sentence is a clue sentence. Without strong correlation, such interpretability is not faithful enough; secondly, although the graph structure is theoretically effective for solving multi-hop question answering problems, recent experiments have proved that the existing graph neural network is only a special attention The force mechanism is not necessary for multi-hop question answering questions. As long as the same additional adjacency matrix information is provided, the same or even better results can be achieved by only using the self-attention mechanism of the Transformer network instead of the graph neural network.

Images