Adaptation of Deep Bidirectional Multilingual Transformers for Russian Language

TL;DR

This paper explores adapting multilingual bidirectional transformers for Russian, demonstrating that transfer learning from multilingual models enhances performance and reduces training time for various NLP tasks.

Contribution

It presents methods for adapting multilingual models to Russian, showing improved performance and efficiency over traditional monolingual training approaches.

Findings

Transfer learning from multilingual to monolingual models boosts task performance.

Multilingual initialization reduces training time for Russian language models.

Pre-trained Russian models are now openly available.

Abstract

The paper introduces methods of adaptation of multilingual masked language models for a specific language. Pre-trained bidirectional language models show state-of-the-art performance on a wide range of tasks including reading comprehension, natural language inference, and sentiment analysis. At the moment there are two alternative approaches to train such models: monolingual and multilingual. While language specific models show superior performance, multilingual models allow to perform a transfer from one language to another and solve tasks for different languages simultaneously. This work shows that transfer learning from a multilingual model to monolingual model results in significant growth of performance on such tasks as reading comprehension, paraphrase detection, and sentiment analysis. Furthermore, multilingual initialization of monolingual model substantially reduces training time. Pre-trained models for the Russian language are open sourced.