Paraphrase Generation Method for Machine Translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine translation systems lack sufficient example texts for improved performance, as the collection of such texts is inadequate, limiting the effectiveness of paraphrasing techniques.
Innovation Solution
A paraphrase generation method that divides an original text into fragments based on a predetermined rule and generates paraphrases within an acceptable limit, using a paraphrase acceptability score and linguistic acceptability score to determine the quality of paraphrases, thereby creating multiple paraphrases from a single text.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If paraphrasing is applied to generate more example texts for machine translation, then the quantity of training data increases, but the quality and acceptability of generated paraphrases deteriorates
Solution Approach 1:
The system employs multiple evaluation mechanisms including paraphrase acceptability score evaluation, linguistic acceptability score evaluation, and machine translation evaluation to assess generated paraphrases. This feedback loop ensures that only high-quality paraphrases meeting predetermined thresholds are selected, maintaining quality while generating large quantities of training data
Solution Approach 2:
The system performs preliminary evaluations during the paraphrase generation process itself, assessing linguistic acceptability and translation quality before final selection. This preliminary filtering ensures that only acceptable paraphrases are generated, preventing quality degradation from the outset rather than correcting issues afterward
2Reliability
If multiple evaluation criteria are used to ensure paraphrase quality, then the reliability of generated paraphrases improves, but the complexity of the generation system increases
Solution Approach 1:
The evaluation system is divided into distinct modular components: paraphrase acceptability score evaluation unit, linguistic acceptability score evaluation unit, and machine translation evaluation unit. Each module independently assesses specific aspects of paraphrase quality, making the complex evaluation process manageable and maintainable while ensuring comprehensive quality control
Data Source
AI summary
A paraphrase generation method according to the present disclosure generates one or more paraphrases of an original text by paraphrasing, within an acceptable limit for accepting paraphrasing, one or more of a plurality of fragments included in the original text into another expression in the language of the original text, the plurality of fragments being obtained by dividing the original text in accordance with a predetermined rule.