ConECT Dataset: Overcoming Data Scarcity in Context-Aware E-Commerce MT

2025-06-05

AI Summary

To address word ambiguity and degraded translation quality in e-commerce neural machine translation (NMT) caused by insufficient contextual information, this paper introduces the first multimodal Czech–Polish e-commerce product translation dataset—comprising 11,400 sentence pairs—enriched with images and structured metadata (category paths and image captions). We propose a novel multimodal NMT framework that jointly integrates visual and structured textual signals, featuring a cross-modal context fusion mechanism implemented via a vision-language model and an enhanced Transformer architecture. Experiments demonstrate substantial improvements over strong baselines: +3.2 BLEU and +4.7 COMET scores, confirming that multimodal context effectively mitigates domain-specific term ambiguity. To foster reproducible research, we publicly release both the first multimodal e-commerce translation benchmark dataset and the corresponding codebase, establishing a new paradigm and foundational resource for domain-adapted machine translation.

Technology Category

• Machine Learning: Multimodal Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP

Application Category

• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search

Abstract

Neural Machine Translation (NMT) has improved translation by using Transformer-based models, but it still struggles with word ambiguity and context. This problem is especially important in domain-specific applications, which often have problems with unclear sentences or poor data quality. Our research explores how adding information to models can improve translations in the context of e-commerce data. To this end we create ConECT -- a new Czech-to-Polish e-commerce product translation dataset coupled with images and product metadata consisting of 11,400 sentence pairs. We then investigate and compare different methods that are applicable to context-aware translation. We test a vision-language model (VLM), finding that visual context aids translation quality. Additionally, we explore the incorporation of contextual information into text-to-text models, such as the product's category path or image descriptions. The results of our study demonstrate that the incorporation of contextual information leads to an improvement in the quality of machine translation. We make the new dataset publicly available.

Problem

• Addressing word ambiguity and context in NMT for e-commerce
• Overcoming data scarcity in Czech-to-Polish product translation
• Improving translation quality with visual and metadata context

Innovation

• Created ConECT dataset with images and metadata
• Tested vision-language model for visual context
• Incorporated product category path into models