Bridge Language Alignment for Statistical Machine Translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Statistical machine translation faces challenges in achieving accurate word alignment, particularly when limited parallel data is available for source and target languages, leading to suboptimal translation quality.
Innovation Solution
The use of bridge languages to determine alignments between source and target languages, generating candidate translations, and combining these alignments to improve translation quality by leveraging greater parallel data availability in bridge languages.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If direct alignment is used between source and target languages with limited parallel data, then the translation system is simple and fast, but the alignment accuracy and translation quality deteriorate
Solution Approach 1:
The patent introduces bridge languages as intermediary elements between source and target languages. Instead of directly aligning source and target languages with limited data, the system aligns source language to bridge language and bridge language to target language separately. This intermediary approach leverages abundant parallel data available for bridge languages to improve overall alignment accuracy, resolving the contradiction between alignment precision and system complexity.
2Reliability
If multiple bridge languages are used to generate diverse alignments, then translation quality improves through diverse hypotheses, but computational complexity and processing time increase
Solution Approach 1:
The patent applies partial action by selectively using one or more bridge languages based on data availability and task requirements, rather than exhaustively using all possible bridge languages. This approach generates sufficient diverse hypotheses to improve translation quality while avoiding the excessive computational burden of processing all possible language pairs, thus balancing translation reliability with processing time.
3Measurement precision
If bridge languages with greater parallel data availability are used, then alignment quality improves, but the system requires more language resources and data storage
Solution Approach 1:
The patent makes bridge language resources serve multiple functions: they are used for aligning source language to target language, for generating candidate translations, and for improving overall translation system performance. By making the bridge language corpus multi-functional, the system maximizes the utility of the additional data storage, allowing the same bridge language data to improve multiple aspects of the translation pipeline simultaneously.
Data Source
AI summary
Systems, methods, and computer program products are provided for statistical machine translation. In some implementations a method is provided. The method includes receiving multi-lingual parallel text associating a source language, a target language, and one or more bridge languages, determining an alignment between the source language and the target language using a first bridge language that is distinct from the source language and the target language, and using the determined alignment to generate a candidate translation of an input text in the source language to the target language.