Efficient Character-level Document Classification by Combining Convolution and Recurrent Layers
This work addresses efficient document classification for natural language processing applications, but it is incremental as it builds on existing character-level methods.
The paper tackles document classification by proposing a neural network architecture that combines convolution and recurrent layers for character-level inputs, achieving comparable performance to character-level convolution-only models with significantly fewer parameters on eight large-scale tasks.
Document classification tasks were primarily tackled at word level. Recent research that works with character-level inputs shows several benefits over word-level approaches such as natural incorporation of morphemes and better handling of rare words. We propose a neural network architecture that utilizes both convolution and recurrent layers to efficiently encode character inputs. We validate the proposed model on eight large scale document classification tasks and compare with character-level convolution-only models. It achieves comparable performances with much less parameters.