Molding CNNs for text: non-linear, non-consecutive convolutions

TL;DR

This paper introduces a novel CNN convolution operation for text that uses tensor algebra and non-consecutive n-grams, improving pattern recognition and efficiency in text classification tasks.

Contribution

It proposes a new non-linear, non-consecutive convolution method using low-rank tensors for better text pattern modeling in CNNs.

Findings

Achieves state-of-the-art accuracy in sentiment classification.

Improves training speed over existing methods.

Effectively models non-consecutive word interactions.

Abstract

The success of deep learning often derives from well-chosen operational building blocks. In this work, we revise the temporal convolution operation in CNNs to better adapt it to text processing. Instead of concatenating word representations, we appeal to tensor algebra and use low-rank n-gram tensors to directly exploit interactions between words already at the convolution stage. Moreover, we extend the n-gram convolution to non-consecutive words to recognize patterns with intervening words. Through a combination of low-rank tensors, and pattern weighting, we can efficiently evaluate the resulting convolution operation via dynamic programming. We test the resulting architecture on standard sentiment classification and news categorization tasks. Our model achieves state-of-the-art performance both in terms of accuracy and training speed. For instance, we obtain 51.2% accuracy on the fine-grained sentiment classification task.