Be My Donor. Transfer the NLP Datasets Between the Languages Using LLM

TL;DR

This paper explores using large language models to transfer annotated NLP datasets between languages, specifically English and Russian, to facilitate resource sharing and reduce annotation efforts.

Contribution

It introduces a pipeline utilizing ChatGPT and Llama models to transfer annotations and demonstrates its effectiveness by training models on translated datasets.

Findings

Successful transfer of annotations between English and Russian

Effective baseline models trained on translated datasets

Potential to accelerate NLP development in underresourced languages

Abstract

In this work, we investigated how one can use the LLM to transfer the dataset and its annotation from one language to another. This is crucial since sharing the knowledge between different languages could boost certain underresourced directions in the target language, saving lots of efforts in data annotation or quick prototyping. We experiment with English and Russian pairs translating the DEFT corpus. This corpus contains three layers of annotation dedicated to term-definition pair mining, which is a rare annotation type for Russian. We provide a pipeline for the annotation transferring using ChatGPT3.5-turbo and Llama-3.1-8b as core LLMs. In the end, we train the BERT-based models on the translated dataset to establish a baseline.