Large dataset unlocks cross language patent claim drafting automation
A Large-Scale Dataset for Cross-Lingual Multi-Jurisdiction Patent Drafting from Patent Cooperation Treaty Applications
Computation and LanguageArtificial Intelligence
Summary
Getting patents approved in different countries means rewriting patent claims to fit each country's language and legal style. The authors collected a huge dataset linking patent claims from Japan and the United States to study how claims change across languages and offices. They treated the problem as more than just translation—it's about adapting claims for each legal setting. They show that machine translation only partly solves this, but teaching computers using linked claims from both countries works better at capturing these differences.
What this means in practice
- •For patent drafting teams: Generate draft patent claims suited to a specific country’s language and legal style using linked multi-jurisdictional claim data.
- •For translation software developers: Develop specialized tools that adapt patent claims across languages and legal environments beyond literal translation.
Authors
Trung Phuong Le, Hirofumi Nonaka, Koji Marusaki, Katsuhito Sudoh, Seiya Kawano
Abstract
Securing patent protection across countries requires drafting claims separately for each national patent office, in the relevant language and according to its legal and stylistic conventions. Patent claims can therefore differ across languages and jurisdictions because of translation, target-office drafting practices, and prosecution-stage changes. We formulate cross-lingual patent drafting as generating plausible candidate claim sets for a specified target language, jurisdiction, and prosecution stage from a source claim set in another language. The task models jurisdiction- and stage-specific transformations beyond content-preserving translation. We construct a large-scale Japanese-English dataset by linking Japan Patent Office (JPO) and United States Patent and Trademark Office (USPTO) records through 884,179 Patent Cooperation Treaty (PCT)-derived patent families. Across six released configurations, these families yield 4,067,168 directional claim-set instances; four core configurations form the main benchmark. Commercial machine translation and zero- and few-shot prompting capture part of the task, while supervised adaptation provides the strongest overall baseline, showing that task-specific patterns can be learned from linked cross-jurisdictional claim sets.