Information Generating Apparatus for Language-Specific Search Precision
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing information search systems using machine learning models face reduced search precision when training data includes queries in multiple languages, as they are optimized for total data sets, leading to performance degradation in specific languages.
Innovation Solution
A system that converts search queries from a second language to a first language for processing by a machine learning model, generating training data with the converted query as a feature value and the searched information in the first language as a correct answer label, thereby improving search precision within the first language.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If training data includes queries in multiple languages, then the machine learning model can handle diverse language inputs, but search precision in specific languages deteriorates due to optimization for total data sets
Solution Approach 1:
The patent segments the training data by language, creating separate training datasets for each language (first language and second language). This allows the machine learning model to be trained separately on monolingual data, ensuring high search precision for each specific language while maintaining the ability to handle multiple languages through separate model instances or selective model selection based on query language.
2Productivity
If the machine learning model is optimized for total data sets, then overall performance across all languages improves, but performance in individual languages deteriorates
Solution Approach 1:
The patent applies local quality by training separate machine learning models or using separate training datasets for each language. This ensures that each language-specific model is optimized for its particular language characteristics, search patterns, and terminology, thereby maintaining high search precision for each individual language rather than compromising it for overall multi-language performance.
Data Source
AI summary
A non-transitory computer-readable recording medium stores an information generating program causing a computer to execute processing of: in an information search using a trained machine learning model generated by machine learning using training data in a first language, converting a search condition in a second language different from the first language to the search condition in the first language; searching information in the first language by inputting the search condition converted to the first language to the machine learning model; and generating training data with the search condition converted to the first language as a feature value and the searched information in the first language as a correct answer label.