Reinforced Mnemonic Reader for Machine Reading Comprehension

Abstract

In this paper, we introduce the Reinforced Mnemonic Reader for machine reading comprehension tasks, which enhances previous attentive readers in two aspects. First, a reattention mechanism is proposed to refine current attentions by directly accessing to past attentions that are temporally memorized in a multi-round alignment architecture, so as to avoid the problems of attention redundancy and attention deficiency. Second, a new optimization approach, called dynamic-critical reinforcement learning, is introduced to extend the standard supervised method. It always encourages to predict a more acceptable answer so as to address the convergence suppression problem occurred in traditional reinforcement learning algorithms. Extensive experiments on the Stanford Question Answering Dataset (SQuAD) show that our model achieves state-of-the-art results. Meanwhile, our model outperforms previous systems by over 6% in terms of both Exact Match and F1 metrics on two adversarial SQuAD datasets.

Published in 27th International Joint Conference on Artificial Intelligence (IJCAI), 2018

References in corpus (13)

• Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
• A Deep Reinforced Model for Abstractive Summarization
• Multi-Task Learning Using Uncertainty to Weigh Losses for Scene Geometry and Semantics
• Machine Comprehension Using Match-LSTM and Answer Pointer
• Adversarial Examples for Evaluating Reading Comprehension Systems
• Multi-Perspective Context Matching for Machine Comprehension
• Reading Wikipedia to Answer Open-Domain Questions
• Reward Augmented Maximum Likelihood for Neural Structured Prediction
• DCN+: Mixed Objective and Deep Residual Coattention for Question Answering
• MEMEN: Multi-layer Embedding with Memory Networks for Machine Comprehension
• Exploring Question Understanding and Adaptation in Neural-Network-Based Question Answering
• Stochastic Answer Networks for Machine Reading Comprehension
• Ruminating Reader: Reasoning with Gated Multi-Hop Attention

Cited by in corpus (26)

• Adaptation of Deep Bidirectional Multilingual Transformers for Russian Language
• DiSAN: Directional Self-Attention Network for RNN/CNN-Free Language Understanding
• Bi-Directional Block Self-Attention for Fast and Memory-Efficient Sequence Modeling
• MEMEN: Multi-layer Embedding with Memory Networks for Machine Comprehension
• Multiresolution Graph Attention Networks for Relevance Matching
• U-Net: Machine Reading Comprehension with Unanswerable Questions
• Temporal Self-Attention Network for Medical Concept Embedding
• Stochastic Answer Networks for Machine Reading Comprehension
• Efficient and Robust Question Answering from Minimal Context over Documents
• Phase Conductor on Multi-layered Attentions for Machine Comprehension
• Conditional Self-Attention for Query-based Summarization
• Open Domain Question Answering Using Early Fusion of Knowledge Bases and Text
• Explicit Utilization of General Knowledge in Machine Reading Comprehension
• A Qualitative Comparison of CoQA, SQuAD 2.0 and QuAC
• Cross-Lingual Transfer Learning for Question Answering
• QAInfomax: Learning Robust Question Answering System by Mutual Information Maximization
• HAS-QA: Hierarchical Answer Spans Model for Open-domain Question Answering
• Enhancing Answer Boundary Detection for Multilingual Machine Reading Comprehension
• Self-Attention Enhanced Patient Journey Understanding in Healthcare System
• Hindi Question Generation Using Dependency Structures
• Multi-Mention Learning for Reading Comprehension with Neural Cascades
• Cross-lingual Machine Reading Comprehension with Language Branch Knowledge Distillation
• Neural Task Representations as Weak Supervision for Model Agnostic Cross-Lingual Transfer
• Phrase-Indexed Question Answering: A New Challenge for Scalable Document Comprehension
• A Study of the Tasks and Models in Machine Reading Comprehension
• BiteNet: Bidirectional Temporal Encoder Network to Predict Medical Outcomes