Sentence Mapping via Back-Translation and Similarity Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing sentence mapping methods face performance degradation when dealing with documents or languages with significant linguistic differences, and are often slow or dependent on the quality of lexical rules and translation performance.
Innovation Solution
A processor-implemented sentence mapping method that translates a target language document into the source language, compares source and translated sentences to determine similarities based on common words, importance levels, and positional relationships, and integrates sentences to form single mappings, thereby improving mapping performance and reducing circular dependencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a lexicon-based sentence mapping method is used, then mapping can be performed using predefined lexical rules, but the mapping speed becomes extremely slow
Solution Approach 1:
The patent introduces a back-translation intermediary process where the target language document is translated back to the source language, and then similarity is determined between the original source language sentences and the back-translated sentences. This intermediary approach avoids the need for slow lexicon-based comparison while maintaining mapping accuracy by leveraging the semantic equivalence preserved through back-translation.
2Reliability
If a machine translation-based sentence mapping method is used, then sentence similarity can be discovered, but circular dependency occurs and performance depends on translator quality
Solution Approach 1:
The patent inverts the traditional machine translation-based approach by translating the target language document back to the source language instead of translating source language to target language. This inversion breaks the circular dependency because the back-translation is performed on an already-translated document, creating a reference frame that is independent of the forward translation quality. The similarity determination then compares the original source sentences with these back-translated versions.
3Productivity
If length-based sentence mapping is used, then sentences can be mapped by comparing word or letter counts, but performance degrades with atypical documents or languages with large linguistic differences
Solution Approach 1:
The patent changes the comparison parameter from superficial features (word count, letter count) to semantic features preserved through back-translation. By comparing the original source language sentences with back-translated sentences in the same language, the method captures semantic equivalence while maintaining efficiency. This parameter change from structural to semantic comparison resolves the degradation issue with atypical documents and linguistically diverse languages.
Data Source
AI summary
A sentence mapping method includes obtaining a source language document in a source language and a target language document in a target language, wherein the target language document is a translation of the source language document, generating a translated document by translating the target language document into the source language, and mapping source language sentences in the source language document and target language sentences with the target language document by comparing the source language document and the translated document.