Joint Translation and Partitioning Model for Long Sentence Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine translation methods fail to accurately translate long sentences due to their partitioning based on punctuation marks, leading to inaccuracies in translation results.
Innovation Solution
A machine translation method and apparatus that utilize a trained model including a partitioning model and a translation model, where sentences are partitioned into blocks based on word vectors to improve translation accuracy by dividing sentences into smaller, more manageable units for translation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sentences are partitioned based on punctuation marks, then translation can be performed on divided clauses, but translation accuracy deteriorates when clauses are relatively long
Solution Approach 1:
The patent applies segmentation by dividing sentences into multiple clauses based on punctuation marks, and then further dividing long clauses into smaller blocks. This multi-level segmentation allows the system to handle both short and long clauses effectively, maintaining translation accuracy while preserving processing capability.
Solution Approach 2:
The patent introduces a new dimension for sentence partitioning by using both punctuation marks and word vectors. Instead of relying solely on punctuation-based segmentation, the system adds a semantic dimension through word vectors to identify meaningful blocks within clauses, thereby improving translation accuracy for long clauses without losing processing efficiency.
2Measurement precision
If sentences are divided into smaller blocks for translation, then translation accuracy improves, but processing complexity increases
Solution Approach 1:
The patent uses segmentation to divide sentences into clauses and then into blocks, which improves translation accuracy by handling smaller units. The dual-level segmentation structure manages complexity by organizing the division process into systematic stages rather than arbitrary fragmentation.
Solution Approach 2:
The trained model serves multiple functions: it performs both sentence partitioning into clauses and block division within clauses. This multi-functionality reduces overall system complexity by consolidating multiple processing tasks into a single unified model rather than requiring separate systems for each function.
3Ease of operation
If traditional punctuation-based partitioning is used, then processing is simple, but translation accuracy deteriorates for long sentences
Solution Approach 1:
The system performs preliminary action by training a model offline that learns both punctuation-based clause boundaries and semantic block structures. This pre-trained model enables the system to maintain simple processing during actual translation while achieving high accuracy, as the complex partitioning logic has already been established during training.
Solution Approach 2:
The trained model acts as an intermediary between simple punctuation-based partitioning and complex semantic understanding. It takes easy-to-obtain punctuation marks as input and produces optimized block divisions that consider semantic meaning, thereby bridging the gap between processing simplicity and translation accuracy.
Data Source
AI summary
A machine translation method includes: receiving a sentence, the sentence including a plurality of words; calling a machine translation model obtained through training, the machine translation model including a partitioning model and a translation model; partitioning the sentence based on the partitioning model and according to word vectors of the words, to obtain to-be-translated blocks, each to-be-translated block including at least one of the words; and translating the sentence based on the translation model and the to-be-translated blocks, to obtain a translation result.