Translation Game System for Parallel Corpus Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional statistical machine translation systems rely heavily on human labor for bilingual data translation, which is time-consuming and costly, and often lacks sufficient bilingual speakers for exotic languages, hindering the development of parallel corpora.
Innovation Solution
A method and apparatus that utilize a translation game played by multilingual individuals to iteratively converge on a final translated structure, using feedback and player models to generate high-quality parallel corpora, making the process fun and cost-effective.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If manually translated bilingual data is used for SMT system development, then translation quality can be maintained, but the process becomes extremely time-consuming and costly
Solution Approach 1:
The system enables self-service translation through iterative gameplay where players automatically generate and refine translations without requiring professional translator intervention. Players compete to converge on accurate translations through the game mechanism, making the system self-sufficient for data collection
Solution Approach 2:
The system implements feedback loops where player translations are compared against reference translations and previous player translations. This feedback mechanism guides players to improve their translations iteratively, ensuring quality improvement over multiple game rounds without manual correction
2Manufacturing precision
If manually translated bilingual data is used for SMT system development, then accurate translation knowledge can be acquired, but the cost increases significantly
Solution Approach 1:
The system replaces expensive professional translators with ordinary players who contribute translations as part of gameplay. These player translations serve as temporary, disposable data points that are refined through iteration, eliminating the need for costly human expert labor while maintaining accuracy through the convergence mechanism
3Quantity of substance
If conventional translation methods are used, then sufficient translation data can be collected, but it takes too long to develop parallel corpora
Solution Approach 1:
The system dynamically scales data collection by allowing multiple players to work on the same translation tasks simultaneously. The iterative convergence mechanism adapts to player performance, automatically adjusting the number of rounds needed, thereby accelerating data collection while maintaining quality through parallel processing of translation tasks
4Quantity of substance
If conventional methods are used for exotic languages, then translation data can be collected, but it becomes difficult to find sufficient bilingual speakers
Solution Approach 1:
The system makes ordinary players universal translators by having them translate from their native language to a target language through the game mechanism. This eliminates the need for rare bilingual speakers, as any player can contribute to exotic language pairs by translating from their available language knowledge, greatly expanding language pair coverage
Data Source
AI summary
A method of generating a statistical machine translation database through a game in which a monolingual structure is provided to a plurality of players. A first translation attempt is received from each of the plurality of players. The first translation attempt from each of the plurality of players is compared. Feedback is provided to each of the plurality of players and the attempts are received and compared to provide feedback to iteratively converge subsequent translations from each of the plurality of players into a final translated structure.