Simultaneous Translation Control Using Phrase Summarization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing simultaneous interpretation technologies face issues with excessive delays or awkward pauses due to significant differences between the input and output time-points of original and translated texts, which hinder smooth communication.
Innovation Solution
An electronic device employs a processor to determine the extent of delay in translation by comparing time-points of phrases, and uses a summary model to generate summary texts for phrases with excessive delay, outputting translated texts after the initial translation is complete, thereby addressing delays and pauses.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If translation is performed in real-time with minimal processing time, then the output time-point of translated text approaches the input time-point of original text, but translation quality and accuracy deteriorate
Solution Approach 1:
The patent segments the original text into multiple phrases and processes them in parallel using multiple translation models with different specializations. This allows simultaneous processing of multiple text segments, reducing overall translation delay while maintaining quality through diverse model perspectives.
Solution Approach 2:
The patent merges the outputs from multiple translation models through a selection mechanism that chooses the most appropriate translation based on various criteria. This combination approach leverages the strengths of different models to produce high-quality translations faster than single-model sequential processing.
2Manufacturing precision
If translation processing is thorough and comprehensive, then translation quality improves, but the output time-point is significantly delayed relative to the input time-point
Solution Approach 1:
The patent performs preliminary actions by pre-processing the original text into multiple phrases and preparing multiple translation models in advance. This preliminary segmentation and model preparation enables faster real-time translation processing without compromising quality, as the heavy lifting is done before the actual translation request.
Solution Approach 2:
The patent implements a dynamic translation system that adaptively selects which translation models to use and how to process different phrases based on real-time conditions. This dynamic approach optimizes the balance between translation quality and speed by adjusting processing depth and model selection according to specific translation needs.
3Manufacturing precision
If multiple translation models are used to improve translation quality, then translation accuracy improves, but device complexity increases
Solution Approach 1:
The patent segments the translation task into multiple parallel processing streams, each handling different phrases with specialized models. This segmentation distributes the complexity across independent modules rather than requiring one complex sequential system, making the overall system more manageable and efficient.
Solution Approach 2:
The patent introduces a selection mechanism as an intermediary that manages the outputs from multiple translation models. This intermediary component coordinates the complex interactions between models, selecting and combining results in a systematic way that reduces overall system complexity while maintaining high translation accuracy.
Data Source
AI summary
An electronic device which obtains a first translated text in which a first phrase in an original text sequence of a first language is translated in a second language by inputting the first phrase to a translation model, determines an extent of delay in translation of the first phrase by comparing a time-point of the first phrase with a time-point of the first translated text, identifies one phrase from among the first phrase and a second phrase following the first phrase based on the extent of delay, obtains a summary text of the identified phrase by inputting the identified phrase to a summary model, obtains a second translated text in which identified phrase or the summary text is translated in the second language by inputting the identified phrase or the summary text to the translation model, and outputs the second translated text after the first translated text is output.