HERIT adapts translation models for Hanja-to-English historical archives
Original title: HERIT: Enabling Global Access to Hanja-to-English Historical Translation by Mitigating Temporal Bias
Authors use RAG on Hanja-Korean corpora to build pseudo-labels, cut temporal bias, and report gains over baselines.
Researchers introduce HERIT, a domain-adaptation pipeline for translating historical Hanja into English. It uses retrieval-augmented generation to create pseudo-labeled data from large Hanja-Korean corpora, expanding scarce human labels and reducing overfitting to particular periods. The authors say automated and human evaluations show consistent gains over baselines. They also apply the model to previously untranslated archive portions.
Unconfirmed: Outperformance claims come from the authors' own automated and human evaluations.
- HERIT targets Hanja-to-English translation of historical archives.
- RAG produces pseudo-labels from abundant Hanja-Korean corpora.
- The pipeline aims to reduce data scarcity and temporal overfitting.
- Authors translated previously untranslated archive portions with the model.
From the original article
1 min read at huggingface.co
Historical archives are invaluable for multidisciplinary research but remain inaccessible globally due to severe language barriers. While manual translation is prohibitively expensive and time-consuming, simply applying existing machine translation models is inadequate due to unique linguistic nuances and the temporal bias inherent in scarce human-labeled corpora.
By Dayeon Jeong, Kyeongpil Kang · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.