Edit Tag Estimation for Post-Editing Guidance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing word-level quality estimation techniques cannot determine which part of the source sentence an error in the machine-translated sentence corresponds to, making it difficult for post-editors to decide what corrections are needed.
Innovation Solution
The technique involves an estimation apparatus that receives extended word alignment and translation quality tags between a source sentence and a translated sentence, and estimates edit tags such as insert, delete, and replace to guide post-editing operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If word-level quality tags are used to identify errors, then error detection capability is improved, but the ability to determine corresponding source sentence parts deteriorates
Solution Approach 1:
The patent segments the translation quality assessment into multiple granular dimensions: source sentence quality tags, translated sentence quality tags, and gap quality tags. Each segment is independently tagged and analyzed, allowing precise identification of error locations and their corresponding source parts without information loss.
Solution Approach 2:
The patent introduces an extended word alignment model as an intermediary that maps between source sentence words and translated sentence words. This intermediary structure preserves the correspondence information, enabling traceability from detected errors back to their source sentence origins.
2Measurement precision
If detailed quality tags are provided for each word, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent employs a unified deep learning model that performs multiple functions simultaneously: extended word alignment, source sentence quality estimation, translated sentence quality estimation, and gap quality estimation. This multi-functional approach achieves high measurement precision without proportionally increasing system complexity.
Solution Approach 2:
The patent changes the parameter representation from simple binary quality tags to a comprehensive multi-dimensional tagging system with extended word alignment information. This parameter transformation enables precise quality measurement while the integrated model structure prevents exponential complexity increase.
Data Source
AI summary
An estimation apparatus includes: an input part configured to receive, as inputs, an extended word alignment between a source sentence and a translated sentence, and translation quality tags between the source sentence and the translated sentence; and an edit tag estimation part configured to estimate an edit tag based on the extended word alignment and the translation quality tags.