Translation Rule Generation Using Document Relevance Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information processing systems face challenges in accurately translating documents due to the lack of relevance-based adaptation of translation rules across different documents, leading to inconsistencies and reduced translation quality.
Innovation Solution
An information processing apparatus that includes a first reception unit for receiving user modifications to translation results, a generation unit for generating translation rules, and a utilization unit that applies these rules based on the relevance between documents, utilizing a hierarchical structure of attributes and score-based determination to ensure appropriate rule application.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If translation rules are customized for specific documents or attributes, then translation accuracy is improved, but device complexity and data requirements increase
Solution Approach 1:
The system pre-calculates and stores relevance scores between documents and between translation rules and attributes before actual translation occurs. This preliminary computation enables rapid, accurate rule selection during translation without complex real-time analysis, resolving the contradiction between translation accuracy and system complexity.
Solution Approach 2:
The patent introduces relevance scores as an intermediary metric that mediates between document attributes and translation rules. Instead of directly matching documents to rules, the system uses pre-computed relevance scores to determine appropriate rule application, simplifying the translation decision process while maintaining accuracy.
2Reliability
If translation rules are applied based on document relevance, then translation quality is improved, but loss of information increases due to selective rule application
Solution Approach 1:
The system changes the parameter of rule application from binary (apply/don't apply) to gradient-based (relevance score thresholding). By adjusting threshold parameters, the system can control the degree of selective rule application, maintaining translation quality while minimizing information loss through configurable parameter tuning.
3Measurement precision
If manual modification of translation results is allowed, then translation accuracy is improved, but loss of time increases due to user intervention
Solution Approach 1:
The system implements feedback by learning from user modifications to translation results. When users correct translation errors, the system incorporates these corrections into updated translation rules, improving future translation accuracy automatically. This feedback mechanism reduces the need for repeated manual interventions over time.
Solution Approach 2:
The system performs self-improvement by automatically generating updated translation rules based on user feedback and relevance analysis. Instead of requiring continuous manual rule creation, the system serves itself by learning from usage patterns and automatically adapting its translation rules, reducing time loss from manual intervention.
Data Source
AI summary
Provided is an information processing apparatus including a first reception unit that receives modification to a translation result of at least one first document from a user, a generation unit that generates a translation rule corresponding to the modification received by the first reception unit, a second reception unit that receives original texts of at least one second document, and a utilization unit that utilizes the translation rule generated by the generation unit at the time of translating the original texts received by the second reception unit, depending on relevance between the at least one first document and the second document.