Enhancing Entity Aware Machine Translation with Multi-task Learning

2025-06-23

AI Summary

Entity-Aware Machine Translation (EAMT) faces two key challenges: scarcity of entity-specific translation data and difficulty in modeling contextual dependencies for entities. To address these, this paper proposes an end-to-end multi-task learning framework that jointly optimizes Named Entity Recognition (NER) and Neural Machine Translation (NMT), leveraging shared semantic representations to enable coordinated modeling of entity boundary detection and translation. Its main contributions are: (1) an entity-aware attention mechanism that explicitly aligns source-side entities with their target-side translations; and (2) an entity consistency loss that enforces cross-task representation alignment. Evaluated on the SemEval 2025 Task 2 benchmark, the method achieves substantial improvements—+4.2 BLEU-E on entity translation accuracy and +2.8 BLEU on overall translation quality—demonstrating both the effectiveness and generalizability of joint modeling for EAMT.

Technology Category

• Natural Language Processing: Machine Translation, Multilinguality, Cross-Lingual NLP
• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Multiagent Learning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Entity-aware machine translation (EAMT) is a complicated task in natural language processing due to not only the shortage of translation data related to the entities needed to translate but also the complexity in the context needed to process while translating those entities. In this paper, we propose a method that applies multi-task learning to optimize the performance of the two subtasks named entity recognition and machine translation, which improves the final performance of the Entity-aware machine translation task. The result and analysis are performed on the dataset provided by the organizer of Task 2 of the SemEval 2025 competition.

Problem

• Addressing data shortage in entity-aware machine translation
• Optimizing entity recognition and translation via multi-task learning
• Improving EAMT performance using SemEval 2025 dataset

Innovation

• Multi-task learning for entity-aware translation
• Combines named entity recognition and translation
• Optimizes performance on SemEval 2025 dataset