Ordered Neurons: Integrating Tree Structures into Recurrent Neural Networks

TL;DR

The paper introduces ON-LSTM, a novel recurrent neural network architecture that explicitly models hierarchical structures in language by ordering neurons, improving performance across multiple language understanding tasks.

Contribution

It proposes the ON-LSTM architecture that incorporates an ordering of neurons to better capture hierarchical language structures, a novel inductive bias for RNNs.

Findings

ON-LSTM outperforms standard LSTM on language modeling tasks.

It achieves competitive results in unsupervised parsing and syntactic evaluation.

The architecture effectively models hierarchical language structures.

Abstract

Natural language is hierarchically structured: smaller units (e.g., phrases) are nested within larger units (e.g., clauses). When a larger constituent ends, all of the smaller constituents that are nested within it must also be closed. While the standard LSTM architecture allows different neurons to track information at different time scales, it does not have an explicit bias towards modeling a hierarchy of constituents. This paper proposes to add such an inductive bias by ordering the neurons; a vector of master input and forget gates ensures that when a given neuron is updated, all the neurons that follow it in the ordering are also updated. Our novel recurrent architecture, ordered neurons LSTM (ON-LSTM), achieves good performance on four different tasks: language modeling, unsupervised parsing, targeted syntactic evaluation, and logical inference.