Tying Word Vectors and Word Classifiers: A Loss Framework for Language Modeling
This addresses a bottleneck in language modeling for NLP researchers, offering a novel method to improve efficiency and performance.
The paper tackles inefficiencies in recurrent neural network language models by introducing a theoretical framework that ties input embedding and output projection matrices, reducing trainable parameters and achieving state-of-the-art performance on the Penn Treebank.
Recurrent neural networks have been very successful at predicting sequences of words in tasks such as language modeling. However, all such models are based on the conventional classification framework, where the model is trained against one-hot targets, and each word is represented both as an input and as an output in isolation. This causes inefficiencies in learning both in terms of utilizing all of the information and in terms of the number of parameters needed to train. We introduce a novel theoretical framework that facilitates better learning in language modeling, and show that our framework leads to tying together the input embedding and the output projection matrices, greatly reducing the number of trainable variables. Our framework leads to state of the art performance on the Penn Treebank with a variety of network models.