Multi-stage attention answer selection method fusing semantics and question key information

Abstract

The invention discloses a multi-stage attention answer selection method fusing semantics and question key information. The multi-stage attention answer selection method comprises two stages, wherein the first stage comprises the steps of obtaining semantic representations bidirectionally output by candidate answers LSTM, performing attention weighted updating on the semantic representations of thecandidate answers by utilizing key information of questions, performing relevancy calculation on the weighted updated semantic representations of the candidate answers and the semantic representations of the questions, and screening out the candidate answers with the highest relevancy; and the second stage is as follows: obtaining semantic representations bidirectionally output by the screened candidate answers LSTM, performing attention weighted updating on the semantic representations of the candidate answers by utilizing semantic information of the questions, performing relevancy calculation on the weighted updated semantic representations of the candidate answers and the semantic representations of the questions again, and selecting an optimal answer from the candidate answers. According to the multi-stage attention answer selection method fusing semantics and question key information, the capturing capability of the model for the candidate answer key information can be enhanced,so that the answer selection accuracy is improved.

Description

technical field [0001] The invention belongs to the technical field of automatic question answering, and in particular relates to a multi-stage attention answer selection method which fuses semantics and key information of questions. Background technique [0002] With the rapid development of Internet technology, the amount of text information in the network has increased exponentially, and it has become an important source for people to obtain information. Therefore, using search engines to retrieve the required information from massive information has become the main way for people to obtain information. Way. However, most of the retrieval strategies of existing search engines are based on string matching, which lacks the ability to mine knowledge from a semantic perspective, resulting in poor accuracy and high redundancy of search results, and requires users to search further from large-scale search results. Only by understanding and screening can we obtain the informati...

AI Technical Summary

Problems solved by technology

In addition, the existing attention-based answer selection models often model the question and answer at the same stage, which is not easy to capture the answer for the answer selection task of selecting a best answer from multiple candidate answers difference between each other

Images