Abstract
Some automated translation systems may be limited by producing a single, high-probability output, which can lack the linguistic diversity beneficial for certain applications. A hybrid, multi-stage workflow can generate multiple contextual translation variants by utilizing a probabilistic generative model, such as a large language model, to produce an initial set of translation candidates based on specified constraints. These candidates may then be programmatically refined, rated for quality, and deduplicated using deterministic and semantic techniques. By producing a curated list of distinct translation options instead of one output, this process can provide broader linguistic coverage to improve the performance of dependent systems, such as content moderation classifiers or specialized search engines.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Kuligin, Leonid and Zettl, Thomas, "A Hybrid System for Multi-Stage Generation of Contextual Translation Variants", Technical Disclosure Commons, (April 01, 2026)
https://www.tdcommons.org/dpubs_series/9700