QED

QED 是一个用于问答解释的语言原则框架。如论文中所述，给定一个问题和一个段落，QED 将答案的解释表示为离散的、人类可解释的步骤的组合：_x000D_
_x000D_
句子选择：= 识别暗示问题答案的句子_x000D_
指称相等 := 识别问题和答案句子中指代同一事物的名词短语_x000D_
谓词蕴含：=确认一旦指称等式被抽象掉，句子中的谓词就蕴含问题中的谓词。

QED is a principled linguistic framework for question answering explanation. As described in the paper, given a question and a paragraph, QED represents the explanation of an answer as a combination of discrete, human-interpretable steps:
Sentence Selection: Identify the sentences that imply the answer to the question
Referential Equality: Identify noun phrases in the question and the answer sentences that refer to the same entity
Predicate Entailment: Confirm that once referential equalities are abstracted away, the predicates in the sentences entail the predicates in the question.

搜集汇总

背景与挑战

背景概述

QED是一个问答解释框架，它将答案解释分解为句子选择、指称相等和谓词蕴含三个步骤。该数据集由Google、斯坦福大学等机构于2020年发布，用于支持自然语言处理中的视觉问答任务。