Partially Shuffling the Training Data to Improve Language Models
Published in
arXiv.org(2019)External Links:
TL
TL;DR
This paper presents a method that partially shuffles the training data between epochs, which makes each batch random, while keeping most sentence ordering intact, and achieves new state of the art results on word-level language modeling on both the Penn Treebank and WikiText-2 datasets.
Abstract
Although SGD requires shuffling the training data between epochs, currently none of the word-level language modeling systems do this. Naively shuffling all sentences in the training data would not permit the model to learn inter-sentence dependencies. Here we present a method that partially shuffles the training data between epochs. This method makes each batch random, while keeping most sentence ordering intact. It achieves new state of the art results on word-level language modeling on both the Penn Treebank and WikiText-2 datasets.
References5 items
1
Direct Output Connection for a High-Rank Language Model
2
Breaking the Softmax Bottleneck: A High-Rank RNN Language Model
3
Regularizing and Optimizing LSTM Language Models
4
Pointer Sentinel Mixture Models
5