Joint Embedding of Words and Labels for Text Classification

TL;DR

This paper introduces a joint embedding approach for text classification that aligns words and labels in the same space, using an attention mechanism to improve interpretability and performance.

Contribution

It presents a novel label-word joint embedding framework with an attention mechanism, enhancing interpretability and outperforming existing methods in accuracy and speed.

Findings

Outperforms state-of-the-art methods in accuracy.

Achieves faster classification speeds.

Maintains interpretability of embeddings.

Abstract

Word embeddings are effective intermediate representations for capturing semantic regularities between words, when learning the representations of text sequences. We propose to view text classification as a label-word joint embedding problem: each label is embedded in the same space with the word vectors. We introduce an attention framework that measures the compatibility of embeddings between text sequences and labels. The attention is learned on a training set of labeled samples to ensure that, given a text sequence, the relevant words are weighted higher than the irrelevant ones. Our method maintains the interpretability of word embeddings, and enjoys a built-in ability to leverage alternative sources of information, in addition to input text sequences. Extensive results on the several large text datasets show that the proposed framework outperforms the state-of-the-art methods by a large margin, in terms of both accuracy and speed.