Towards Human-level Machine Reading Comprehension: Reasoning and Inference with Multiple Strategies

By Yichong Xu, Jingjing Liu, Jianfeng Gao, Yelong Shen, Xiaodong Liu. 2017

29 citations recorded by the source corpus.

This paper presents a new MRC model that is capable of three key comprehension skills: 1) handling rich variations in question types; 2) understanding potential answer choices; and 3) drawing inference through multiple sentences. The model is based on the proposed MUlti-Strategy Inference for Comprehension (MUSIC) architecture, which is able to dynamically apply different attention strategies to different types of questions on the fly. By incorporating a multi-step inference engine analogous to ReasoNet (Shen et al., 2017), MUSIC can also effectively perform multi-sentence inference in generating answers. Evaluation on the RACE dataset shows that the proposed method significantly outperforms previous state-of-the-art models by 7.5% in relative accuracy.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Yi Tay
2. Mo Yu
3. Anh Tuan Luu
4. Siu Cheung Hui
5. Shuohang Wang
6. Shiyu Chang
7. Jing Jiang
8. Yiming Cui

In Computer Vision and Pattern Recognition

1. Shuohang Wang
2. Mo Yu
3. Jing Jiang
4. Shiyu Chang
5. Yiming Cui
6. Shijin Wang
7. Guoping Hu
8. Yi Tay

In Natural Language Processing

1. Yi Tay
2. Mo Yu
3. Shuohang Wang
4. Anh Tuan Luu
5. Siu Cheung Hui
6. Shiyu Chang
7. Jing Jiang
8. Yiming Cui

Citations in the corpus

• Option Comparison Network for Multiple-choice Reading Comprehension (2019, background): “…models have been proposed for MCRC and achieve promising re-sults in recent years (Chen et al., 2016; Yin et al., 2016; Trischler et al., 2016; Dhingra et al., 2017; Tay et al., 2018; Parikh et al., 2018; Zhu et al., 2018; Wang et al., 2018; Xu et al., 2017; Sun et al., 2018; Zhang et al., 2019).”
• A Co-Matching Model for Multi-choice Reading Comprehension (2018, methodology): “• MUSIC (Xu et al., 2017) integrates different sequence matching strategies into the model and also adds a unit of multi-step reasoning for selecting the answer.”
• Improving Question Answering by Commonsense-Based Pre-Training (2018, unknown): “With the revival of interest in neural network approaches, current top-performing methods in MRC datasets are dominated by neural models (Xu et al. 2017; Yu et al. 2018), some of which even achieve human-level performance on several particular datasets.”
• Convolutional Spatial Attention Model for Reading Comprehension with Multiple-Choice Questions (2018, background): “Xu et al.(2017) proposed the Dynamic Fusion Networks (DFN), which uses multi-hop reasoning mechanism for this task.”
• Incorporating Structured Commonsense Knowledge in Story Completion (2018, background): “2018), and dynamic fusion based model like DFN (Xu et al. 2017).”
• Multi-range Reasoning for Machine Comprehension (2018, methodology): “Moreover, our model outperforms DFN by 2% − 6% on the RACE benchmark and other strong baselines such as the Gated Attention Reader by 10% .”
• UvA-DARE (Digital Academic Repository) Learning to Transform, Combine, and Reason in Open-domain Question Answering (unknown): “Multihop reasoning is one of the key requirements for reading comprehension [13, 30, 40].”
• ChatGPT: Jack of all trades, master of none (2023, background): “RACE dataset [86] is a reading comprehension dataset consisting of over 100,000 multiple-choice questions relating to about 28,000 passages from various topics.”
• Multi-Matching Network for Multiple Choice Reading Comprehension (2019, methodology): “Among existing works, DFN (Xu et al. 2017) and Hier-Co-Matching (Wang et al.”