Language Modeling with Highway LSTM

TL;DR

This paper introduces Highway LSTM, an extension of standard LSTM with highway networks inside, which increases depth and improves language modeling performance in speech recognition tasks.

Contribution

The paper proposes Highway LSTM models with highway networks integrated inside LSTM units, enhancing depth and accuracy over traditional LSTM language models.

Findings

HW-LSTM improves speech recognition accuracy.

Achieves new best performance on Switchboard and CallHome tasks.

Abstract

Language models (LMs) based on Long Short Term Memory (LSTM) have shown good gains in many automatic speech recognition tasks. In this paper, we extend an LSTM by adding highway networks inside an LSTM and use the resulting Highway LSTM (HW-LSTM) model for language modeling. The added highway networks increase the depth in the time dimension. Since a typical LSTM has two internal states, a memory cell and a hidden state, we compare various types of HW-LSTM by adding highway networks onto the memory cell and/or the hidden state. Experimental results on English broadcast news and conversational telephone speech recognition show that the proposed HW-LSTM LM improves speech recognition accuracy on top of a strong LSTM LM baseline. We report 5.1% and 9.9% on the Switchboard and CallHome subsets of the Hub5 2000 evaluation, which reaches the best performance numbers reported on these tasks to date.