Science Question Answering using Instructional Materials

TL;DR

This paper introduces a unified max-margin framework that leverages instructional materials to improve elementary science question answering by uncovering hidden answer structures, leading to superior performance over existing methods.

Contribution

The paper presents a novel max-margin learning approach that models hidden answer structures using instructional materials for elementary science questions.

Findings

Outperforms several strong baseline methods

Effectively models hidden answer structures

Achieves higher accuracy on elementary science test questions

Abstract

We provide a solution for elementary science test using instructional materials. We posit that there is a hidden structure that explains the correctness of an answer given the question and instructional materials and present a unified max-margin framework that learns to find these hidden structures (given a corpus of question-answer pairs and instructional materials), and uses what it learns to answer novel elementary science questions. Our evaluation shows that our framework outperforms several strong baselines.