Work

Trivial Transfer Learning for Low-Resource Neural Machine Translation

Year: 2018

Type: conference-paper

Abstract: Transfer learning has been proven as an effective technique for neural machine translation under low-resource conditions.Existing methods require a common target language, language relatedness, or specific training tricks and regimes.We present a simple transfer learning method, where we first train a "parent" model for a high-resource language pair and then continue the training on a lowresource pair only by replacing the training corpus.This "child" model performs significantly better than the baseline trained for lowresource pair only.We are the first to show this for targeting different languages, and we observe the improvements even for unrelated languages with different alphabets.

DOI: 10.18653/v1/w18-6325

Authors: Tom Kocmi, Ondřej Bojar

Institution: Charles University

Language: English

Topic: Natural Language Processing Techniques

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: gold

Funders: Ministerstvo Školství, Mládeže a Tělovýchovy, Grantová Agentura České Republiky

Awards: G12500569, G3711331958, G4131035492, G5641063287, Language Research Infrastructure in the Czech Republic

OpenAlex ID: W2892004290