Network-Efficient Distributed Word2vec Training System for Large Vocabularies

TL;DR

This paper introduces a distributed Word2vec training system capable of efficiently handling vocabularies with hundreds of millions of words, reducing network traffic and maintaining vector quality for large-scale NLP applications.

Contribution

The paper presents a novel distributed training system for Word2vec that significantly improves scalability and efficiency for very large vocabularies compared to existing methods.

Findings

System enables training on 100+ million words with less network traffic.

Vector quality remains comparable to non-distributed training.

Deployed in Yahoo's ad platform, improving business metrics.

Abstract

Word2vec is a popular family of algorithms for unsupervised training of dense vector representations of words on large text corpuses. The resulting vectors have been shown to capture semantic relationships among their corresponding words, and have shown promise in reducing a number of natural language processing (NLP) tasks to mathematical operations on these vectors. While heretofore applications of word2vec have centered around vocabularies with a few million words, wherein the vocabulary is the set of words for which vectors are simultaneously trained, novel applications are emerging in areas outside of NLP with vocabularies comprising several 100 million words. Existing word2vec training systems are impractical for training such large vocabularies as they either require that the vectors of all vocabulary words be stored in the memory of a single server or suffer unacceptable training latency due to massive network data transfer. In this paper, we present a novel distributed, parallel training system that enables unprecedented practical training of vectors for vocabularies with several 100 million words on a shared cluster of commodity servers, using far less network traffic than the existing solutions. We evaluate the proposed system on a benchmark dataset, showing that the quality of vectors does not degrade relative to non-distributed training. Finally, for several quarters, the system has been deployed for the purpose of matching queries to ads in Gemini, the sponsored search advertising platform at Yahoo, resulting in significant improvement of business metrics.