On Training Bi-directional Neural Network Language Model with Noise Contrastive Estimation

TL;DR

This paper explores training bi-directional neural network language models using noise contrastive estimation, showing improvements over traditional methods but not surpassing uni-directional models in a specific task.

Contribution

It introduces a novel training approach for bi-directional neural network language models using noise contrastive estimation.

Findings

NCE-trained bi-directional NNLM outperforms maximum likelihood trained models.

Bi-directional NNLM does not outperform baseline uni-directional NNLM.

The approach shows promise but has limitations in surpassing existing models.

Abstract

We propose to train bi-directional neural network language model(NNLM) with noise contrastive estimation(NCE). Experiments are conducted on a rescore task on the PTB data set. It is shown that NCE-trained bi-directional NNLM outperformed the one trained by conventional maximum likelihood training. But still(regretfully), it did not out-perform the baseline uni-directional NNLM.