Efficient Non-parametric Estimation of Multiple Embeddings per Word in Vector Space

TL;DR

This paper introduces an efficient, scalable extension to the Skip-gram model that learns multiple embeddings per word, capturing polysemy and improving word similarity tasks at large scale.

Contribution

It presents a non-parametric, joint approach to sense discrimination and embedding learning, significantly advancing multi-sense word embedding methods.

Findings

Achieved state-of-the-art results in word similarity in context tasks.

Trained on nearly 1 billion tokens in less than 6 hours on a single machine.

Demonstrated scalability and efficiency of the proposed method.

Abstract

There is rising interest in vector-space word embeddings and their use in NLP, especially given recent methods for their fast estimation at very large scale. Nearly all this work, however, assumes a single vector per word type ignoring polysemy and thus jeopardizing their usefulness for downstream tasks. We present an extension to the Skip-gram model that efficiently learns multiple embeddings per word type. It differs from recent related work by jointly performing word sense discrimination and embedding learning, by non-parametrically estimating the number of senses per word type, and by its efficiency and scalability. We present new state-of-the-art results in the word similarity in context task and demonstrate its scalability by training with one machine on a corpus of nearly 1 billion tokens in less than 6 hours.