Learning label-label correlations in Extreme Multi-label Classification via Label Features

TL;DR

This paper introduces Gandalf, a novel method for extreme multi-label classification that leverages label features and label co-occurrence graphs to generate additional training data, improving tail label prediction without extra computational costs.

Contribution

Gandalf is a new approach that uses label graphs and features to create supplementary training instances, enhancing existing XMC algorithms' performance on large-scale datasets.

Findings

Models trained on augmented data outperform original datasets on PSP@k.

Gandalf achieves an average 5% improvement across multiple algorithms and datasets.

The method is plug-and-play and does not increase computational overhead.

Abstract

Extreme Multi-label Text Classification (XMC) involves learning a classifier that can assign an input with a subset of most relevant labels from millions of label choices. Recent works in this domain have increasingly focused on a symmetric problem setting where both input instances and label features are short-text in nature. Short-text XMC with label features has found numerous applications in areas such as query-to-ad-phrase matching in search ads, title-based product recommendation, prediction of related searches. In this paper, we propose Gandalf, a novel approach which makes use of a label co-occurrence graph to leverage label features as additional data points to supplement the training distribution. By exploiting the characteristics of the short-text XMC problem, it leverages the label features to construct valid training instances, and uses the label graph for generating the corresponding soft-label targets, hence effectively capturing the label-label correlations. Surprisingly, models trained on these new training instances, although being less than half of the original dataset, can outperform models trained on the original dataset, particularly on the PSP@k metric for tail labels. With this insight, we aim to train existing XMC algorithms on both, the original and new training instances, leading to an average 5% relative improvements for 6 state-of-the-art algorithms across 4 benchmark datasets consisting of up to 1.3M labels. Gandalf can be applied in a plug-and-play manner to various methods and thus forwards the state-of-the-art in the domain, without incurring any additional computational overheads.