Meta-Embedding as Auxiliary Task Regularization

TL;DR

This paper introduces a meta-embedding reconstruction as an auxiliary task to improve word and sentence embedding performance, leveraging labeled datasets for regularization and achieving significant gains in similarity and downstream tasks.

Contribution

It proposes a novel auxiliary task of reconstructing ensemble embeddings to regularize meta-embedding learning, enhancing performance across intrinsic and extrinsic evaluations.

Findings

Improved word similarity scores with an average increase of 11.33 in Spearman correlation.

Best performance in 4 out of 6 word similarity datasets using cosine and Brier's loss.

Enhanced downstream task performance in sequence tagging and sentence classification.

Abstract

Word embeddings have been shown to benefit from ensambling several word embedding sources, often carried out using straightforward mathematical operations over the set of word vectors. More recently, self-supervised learning has been used to find a lower-dimensional representation, similar in size to the individual word embeddings within the ensemble. However, these methods do not use the available manually labeled datasets that are often used solely for the purpose of evaluation. We propose to reconstruct an ensemble of word embeddings as an auxiliary task that regularises a main task while both tasks share the learned meta-embedding layer. We carry out intrinsic evaluation (6 word similarity datasets and 3 analogy datasets) and extrinsic evaluation (4 downstream tasks). For intrinsic task evaluation, supervision comes from various labeled word similarity datasets. Our experimental results show that the performance is improved for all word similarity datasets when compared to self-supervised learning methods with a mean increase of $11.33$ in Spearman correlation. Specifically, the proposed method shows the best performance in 4 out of 6 of word similarity datasets when using a cosine reconstruction loss and Brier's word similarity loss. Moreover, improvements are also made when performing word meta-embedding reconstruction in sequence tagging and sentence meta-embedding for sentence classification.