Cross-language search via dynamic machine translation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems fail to effectively search for resources when a query is entered in a language different from the language of the application interface, leading to irrelevant results due to language barriers, especially in dynamic environments where static solutions like lookup tables are impractical.
Innovation Solution
A computing system that receives a query in a second language, translates it, and searches resources accessible by a second application in the same language, returning results in the language of the interface, while also performing additional searches and refining a machine translation model based on user feedback to improve query-resource association.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a query is entered in a language different from the application interface language, then the system can potentially access broader multilingual resources, but the search results become irrelevant due to language mismatch
Solution Approach 1:
The patent introduces a machine translation model as an intermediary component that translates queries from the user's language to the language of the application interface. This mediator enables cross-language searching while maintaining search result relevance by ensuring the translated query matches the resource language, thus resolving the language mismatch problem.
Solution Approach 2:
The system dynamically changes the language parameter of the query by using machine translation to convert the input query into the target language. This parameter transformation allows the search system to process queries in different languages while maintaining accurate matching with resources in the interface language, thereby preserving search relevance across languages.
2Ease of manufacture
If static lookup tables are used for language translation, then implementation is simple, but the system cannot adapt to dynamic environments and new language pairs
Solution Approach 1:
The patent replaces static lookup tables with a dynamic machine translation model that can adapt to new language pairs and evolving language usage. The model is trained on dynamic datasets and can be continuously updated, allowing the system to handle new languages and language variations without requiring complete reimplementation, thus achieving both adaptability and operational simplicity.
Solution Approach 2:
The machine translation model is trained using automated processes that leverage large corpora of parallel texts. The system self-improves by continuously learning from new data sources and adapting to emerging language patterns, reducing the need for manual curation and updating of translation resources while maintaining high adaptability to new language pairs.
3Measurement precision
If machine translation is performed for every query, then cross-language search accuracy improves, but processing time and computational resources increase
Solution Approach 1:
The system performs preliminary actions by pre-translating frequently occurring queries or by preparing translation models in advance for common language pairs. This allows the actual search process to use pre-computed translations, reducing the real-time processing burden while maintaining high translation accuracy for cross-language queries.
Solution Approach 2:
The patent applies machine translation selectively based on the specific query characteristics and language pair requirements. Instead of uniformly translating every query, the system identifies when translation is necessary and applies it locally to those specific cases, optimizing the balance between translation accuracy and processing efficiency by avoiding unnecessary translations for same-language queries or simple keywords.
Data Source
AI summary
In accordance with one disclosed method, a computing system may receive, via a first version of an application presenting an interface in a first language, a first query entered in a second language different from the first language. The computing system may search resources accessible to a second version of the application to identify at least a first resource corresponding to the first query, the second version of the application being in the second language. Based at least in part on a result of the searching, an indication of the first resource may be returned to the interface.