Recurrent Memory Networks for Language Modeling

TL;DR

The paper introduces Recurrent Memory Networks (RMN), a new RNN architecture that enhances language modeling and sentence completion tasks while providing better interpretability of the model's internal mechanisms.

Contribution

The paper proposes RMN, a novel RNN architecture that improves language modeling performance and offers insights into the model's internal functioning and learned patterns.

Findings

RMN outperforms LSTM on language modeling across three datasets.

RMN achieves 69.2% accuracy on Sentence Completion Challenge, surpassing previous state-of-the-art.

RMN captures various linguistic dimensions effectively.

Abstract

Recurrent Neural Networks (RNN) have obtained excellent result in many natural language processing (NLP) tasks. However, understanding and interpreting the source of this success remains a challenge. In this paper, we propose Recurrent Memory Network (RMN), a novel RNN architecture, that not only amplifies the power of RNN but also facilitates our understanding of its internal functioning and allows us to discover underlying patterns in data. We demonstrate the power of RMN on language modeling and sentence completion tasks. On language modeling, RMN outperforms Long Short-Term Memory (LSTM) network on three large German, Italian, and English dataset. Additionally we perform in-depth analysis of various linguistic dimensions that RMN captures. On Sentence Completion Challenge, for which it is essential to capture sentence coherence, our RMN obtains 69.2% accuracy, surpassing the previous state-of-the-art by a large margin.