Character-Aware Neural Language Models

TLDR

The paper proposes a character‑only neural language model that predicts words. The model composes character embeddings with a CNN and highway network, then feeds them into an LSTM language model. On the Penn Treebank it matches state‑of‑the‑art performance with 60 % fewer parameters, and on morphologically rich languages it outperforms word‑ and morpheme‑level LSTM baselines, showing that character inputs alone can capture semantic and orthographic information.

Abstract

We describe a simple neural language model that relies only on character-level inputs. Predictions are still made at the word-level. Our model employs a convolutional neural network (CNN) and a highway net work over characters, whose output is given to a long short-term memory (LSTM) recurrent neural network language model (RNN-LM). On the English Penn Treebank the model is on par with the existing state-of-the-art despite having 60% fewer parameters. On languages with rich morphology (Arabic, Czech, French, German, Spanish, Russian), the model outperforms word-level/morpheme-level LSTM baselines, again with fewer parameters. The results suggest that on many languages, character inputs are sufficient for language modeling. Analysis of word representations obtained from the character composition part of the model reveals that the model is able to encode, from characters only, both semantic and orthographic information.