AI Translation Co-occurrence Word Selection for Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine translation models, particularly non-autoregressive models, face challenges in translation accuracy due to their inability to effectively fuse multiple translation models and manage dependency relationships between tokens, leading to inefficient decoding processes and incorrect inference results.
Innovation Solution
The proposed method involves acquiring intermediate translation results from multiple pre-trained translation models in each iteration, identifying co-occurrence words across these results, and using these words to determine the target translation result, thereby improving translation accuracy by avoiding confusion from non-co-occurrence words.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple pre-trained translation models are used to generate intermediate translation results, then translation accuracy can be improved, but the complexity of the translation system increases
Solution Approach 1:
The patent combines multiple pre-trained translation models into a unified translation system that processes sentences through all models simultaneously. The intermediate translation results from each model are merged and analyzed to identify co-occurrence words, creating a collective decision-making mechanism that improves accuracy while managing complexity through systematic integration.
Solution Approach 2:
The patent introduces an intermediary processing layer that analyzes intermediate translation results from multiple models. This layer identifies co-occurrence words across different model outputs and uses them as mediators to determine the final translation result, thereby managing the complexity of multiple models through a structured intermediate analysis stage.
2Reliability
If co-occurrence words are identified from multiple intermediate translation results, then translation reliability is improved, but the processing time increases
Solution Approach 1:
The patent performs preliminary identification of co-occurrence words from intermediate translation results during the translation process. By pre-identifying these reliable words before final result generation, the system establishes a foundation for accurate translation that reduces the need for repeated processing and corrections, thereby managing processing time while maintaining reliability.
Solution Approach 2:
The patent implements a feedback mechanism where intermediate translation results are analyzed to identify co-occurrence words, which then inform the determination of the final translation result. This feedback loop allows the system to learn from multiple model outputs and adjust the final translation based on consensus patterns, improving reliability through iterative refinement.
3Measurement precision
If intermediate translation results from multiple models are analyzed, then translation precision is improved, but the computational resources required increase
Solution Approach 1:
The patent extracts only the essential information from intermediate translation results - specifically the co-occurrence words that appear across multiple model outputs. By taking out and focusing on these key consensus elements rather than processing entire translation sequences, the system improves precision through multi-model analysis while reducing computational overhead by analyzing only the critical overlapping information.
Data Source
AI summary
A translation method, an electronic device and a storage medium, which relate to the field of artificial intelligence technologies, such as machine learning technologies, information processing technologies, are disclosed. An implementation includes: acquiring an intermediate translation result generated by each of multiple pre-trained translation models for a to-be-translated specified sentence in a same iteration of a translation process, so as to obtain multiple intermediate translation results; acquiring a co-occurrence word based on the multiple intermediate translation results; and acquiring a target translation result of the specified sentence based on the co-occurrence word.