Method for solving reading comprehension question based on semantic analysis and SMT solution

Abstract

The invention provides a method for solving a reading comprehension question based on semantic analysis and SMT solution. The method comprises the following steps: firstly, performing semantic parsingon the reading comprehension question to generate a first-order logic formula corresponding to the reading comprehension question, then introducing four hypotheses, namely, a unique name hypothesis,a closed world hypothesis, a closure cause hypothesis and a unique answer hypothesis, to generate a first-order logic formula corresponding to candidate answers; and finally, verifying whether a knowledge base contains the first-order logic formula corresponding to the candidate answers by using a SMT solver, and then figuring out the answer. Compared with the existing methods of expressing textsby using neural network models and word vectors, the method provided by the invention has the advantages being able to better establish the relationship between the events described in the reading comprehension question, thereby giving a reading comprehension system a stronger expression ability and reasoning ability.

Description

technical field [0001] The invention relates to the field of computers, in particular to a method for solving reading comprehension problems based on semantic analysis and SMT (Satisfiability Modulo Theories). Background technique [0002] It is a challenging task for a machine to read a piece of text and then answer questions related to that text. A machine can only complete this task if it has the ability to read, comprehend and reason about the text. In order to test the strength of machine reading comprehension and reasoning and solving abilities, scholars have proposed a variety of data sets, including MCTest, WikiQA, SQuAD, MS MARCO, etc. The reading comprehension questions in these datasets all consist of a long text and related questions. Their characteristics are: 1. The text is long, and the machine needs to eliminate the interference of information that is not related to the question in the text; 2. The answers to most questions appear directly in a sentence in ...

AI Technical Summary

Problems solved by technology

However, these methods are not suitable for solving reading comprehension questions that require deep reasoning

Therefore, these methods cannot solve the reading comprehension questions in the WSC data set containing common sense very well, so there is an urgent need for a reading comprehension problem solving method with stronger expressive and reasoning abilities

Images