Language Modeling with Highway LSTM
This work addresses speech recognition accuracy for English broadcast news and conversational telephone speech, representing an incremental improvement over existing LSTM language models.
The paper tackled language modeling for speech recognition by extending LSTM with highway networks to increase depth in the time dimension, resulting in a Highway LSTM model that improved speech recognition accuracy by 5.1% and 9.9% on Switchboard and CallHome subsets, achieving state-of-the-art performance.
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.