Brotherhood at WMT 2024: Leveraging LLM-Generated Contextual Conversations for Cross-Lingual Image Captioning

TL;DR

This paper introduces a novel cross-lingual image captioning method that uses large language models to generate contextual conversations, improving translation quality without traditional training.

Contribution

The authors propose leveraging instruction-tuned prompting of LLMs to create synthetic conversations for cross-lingual captioning, avoiding traditional training or fine-tuning.

Findings

Achieved 37.90 BLEU on English-Hindi challenge set.

Ranked first and second for English-Hausa on leaderboards.

Explored trade-offs between BLEU scores and semantic similarity.

Abstract

In this paper, we describe our system under the team name Brotherhood for the English-to-Lowres Multi-Modal Translation Task. We participate in the multi-modal translation tasks for English-Hindi, English-Hausa, English-Bengali, and English-Malayalam language pairs. We present a method leveraging multi-modal Large Language Models (LLMs), specifically GPT-4o and Claude 3.5 Sonnet, to enhance cross-lingual image captioning without traditional training or fine-tuning. Our approach utilizes instruction-tuned prompting to generate rich, contextual conversations about cropped images, using their English captions as additional context. These synthetic conversations are then translated into the target languages. Finally, we employ a weighted prompting strategy, balancing the original English caption with the translated conversation to generate captions in the target language. This method achieved competitive results, scoring 37.90 BLEU on the English-Hindi Challenge Set and ranking first and second for English-Hausa on the Challenge and Evaluation Leaderboards, respectively. We conduct additional experiments on a subset of 250 images, exploring the trade-offs between BLEU scores and semantic similarity across various weighting schemes.