Translation Support System for Game Narrative Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine translation systems face challenges in maintaining high translation accuracy, especially for narrative texts in games, due to labor-intensive manual interventions and the difficulty in handling multiple languages and ambiguous expressions, while also being cost-inefficient and prone to accuracy fluctuations.
Innovation Solution
A translation support system that reduces manual interventions by using an input unit, error database, controlled-source-language-sentence database, and control unit to classify and convert non-machine-translatable sentences into controlled-source-language sentences, enabling machine translation while maintaining high accuracy and reducing costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If rule base machine translation is used, then translation accuracy is stably high, but labor and cost increase due to manual construction of language rules, grammar rules, and dictionary for each language pair
Solution Approach 1:
The system enables statistical machine translation to automatically generate translation rules and dictionaries from corpus data without requiring manual construction for each language pair. The feedback learning mechanism allows the system to self-improve translation accuracy by learning from user corrections and accumulated translation data, eliminating the need for continuous manual rule creation while maintaining high translation quality
Solution Approach 2:
The system changes the approach from manual rule-based parameters to automated statistical parameters derived from large volumes of translation pairs. By using feedback learning, the system dynamically adjusts translation models based on accumulated data, allowing translation accuracy to improve over time without increasing manual labor
2Ease of manufacture
If statistical machine translation is used, then cost is reduced and feedback learning enables progressive improvement, but translation accuracy fluctuates and requires large corpus preparation
Solution Approach 1:
The system performs preliminary preparation by automatically building translation models and dictionaries from large volumes of translation pairs before actual translation tasks. This pre-processing stage establishes the foundation for accurate translation, reducing the need for manual interventions during operation while maintaining consistent accuracy
Solution Approach 2:
The system implements feedback learning where user corrections and translation results are continuously fed back to improve the translation model. This closed-loop mechanism allows the system to progressively improve translation accuracy over time, stabilizing performance while maintaining low operational costs
3Manufacturing precision
If manual rewriting into controlled natural language is performed, then translation accuracy is improved, but labor and time increase
Solution Approach 1:
The system replaces the mechanical process of manual rewriting with automated statistical machine translation. By using large volumes of translation pairs and feedback learning, the system automatically achieves high translation accuracy without requiring human intervention for rewriting, thereby eliminating the time and labor costs associated with manual controlled natural language conversion
Data Source
AI summary
The present invention is a translation support system for supporting machine translation from a source-language sentence into a target-language sentence, the translation support system including an input unit that accepts an input of a source-language sentence to be translated; an error database that at least stores words or combinations of words included in a plurality of source-language sentences for which machine translation from the source-language sentences into target-language sentences is not performed correctly; a controlled-source-language-sentence database that stores a plurality of source-language sentences as well as controlled source-language sentences, which are source-language sentences that are controlled, corresponding to the plurality of source-language sentences and expressed in a format satisfying predetermined conditions; a control unit that classifies whether or not the input source-language sentence is machine-translatable; and an output unit that is capable of outputting the input source-language sentence classified as being non-machine-translatable.