The Impact of Domain-Specific Terminology on Machine Translation for Finance in European Languages
The study analyzes financial machine translation challenges and terminology accuracy across 22 European languages using multi-parallel datasets.
What it examines
The study introduces challenges in machine translation for finance, focusing on domain-specific terminology in European languages. It presents a new multi-parallel corpus from the ECB and compares MT systems and LLMs using an English financial glossary to assess how accurately financial terms are translated.
What it concludes
The research finds strong links between term accuracy and translation quality. It provides resources and methods useful for improving financial translations. Applications include better regulatory reporting and informed decision-making in finance, while future work can expand domain and language coverage.
Evidence objects
The study reveals that domain-specific financial terminology critically impacts machine translation quality across 22 European languages, showing surprising strong correlations between terminology accuracy and overall performance in specialized financial texts.
key_findings bullet 1 · key_findings · validation V0
Researchers introduce a novel financial corpus and a glossary, paired with a methodology to analyze term match accuracy at corpus and segment levels, outperforming large language models on financial content.
key_findings bullet 2 · key_findings · validation V0
Combining advanced automatic alignment with manually curated glossaries, the study provides statistical analyses revealing that morphological complexity significantly influences translation metrics, while acknowledging limitations focused on macroeconomic texts in practice.
key_findings bullet 3 · key_findings · validation V0
The paper introduces a novel multi-parallel corpus for financial NLP by incorporating $$specialized financial terminology$$ translation across European languages. Its originality arises from a domain-specific approach with an evaluation methodology. Novel insights and resource creation offer fresh perspectives, making it compelling for advancing financial markets analysis and financial modeling research.
key_findings bullet 4 · key_findings · validation V0
Raw abstract and provenance
- … financial glossary, we propose a methodology to analyze the relationship between translation performance (into English) and the accuracy of financial … tilingual financial …
Source row: 1951 · abstract type: snippet