Language Determination for Text-to-Speech with Non-Standard Elements
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing electronic devices struggle to accurately determine the language of text that includes numbers, symbols, or formulas during the text-to-speech conversion process, leading to inaccuracies in speech synthesis.
Innovation Solution
A language determination method and apparatus that involve obtaining a first text and a reply text, determining a target sentence, and identifying the languages of the first text and preceding parts of the target sentence to accurately determine the target language for speech conversion based on the target sentence, reply text, first language, and second language.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the electronic device directly recognizes the language of text output by the language model, then the text-to-speech conversion can be performed, but the language determination accuracy deteriorates when the text includes numbers, sequence numbers, or other non-standard elements
Solution Approach 1:
The reply text is segmented into multiple sentences, and each sentence is further divided into word segments. This segmentation allows the system to analyze and determine the language of each individual word or segment independently, rather than treating the entire text as a single unit. This is particularly effective for texts containing numbers, sequence numbers, and formulas, as each element can be evaluated in its specific contextual position.
Solution Approach 2:
The patent applies different language determination strategies to different parts of the text based on their local characteristics. For standard words, the system uses language recognition models, while for non-standard elements like numbers and formulas, the system determines language based on contextual clues from surrounding words and the overall text structure. This localized approach allows the system to handle diverse text elements with appropriate methods for each.
2Measurement precision
If the electronic device uses a simple language recognition method, then the processing speed is fast, but the accuracy deteriorates for complex text structures containing multiple languages and formats
Solution Approach 1:
The system performs preliminary actions by first segmenting the text into sentences and then into word segments before language determination. It also pre-identifies non-standard elements such as numbers, formulas, and special characters. This preliminary processing organizes the complex text structure into manageable units, making the subsequent language determination more accurate and systematic without requiring overly complex real-time processing.
Solution Approach 2:
The language determination process is dynamic and adaptive. The system adjusts its determination strategy based on the type of text element being analyzed. For standard words, it uses language recognition models; for non-standard elements, it uses contextual analysis. This dynamic approach allows the system to handle varying text complexities with appropriate levels of processing depth.
Data Source
AI summary
The present disclosure provides a language determination method and electronic device. The method includes: obtaining a first text and a reply text for the first text; determining a target sentence in the reply text; determining a first language of the first text and a second language of a preceding part of the target sentence; and determining a target language used when the target sentence is converted into speech based on the target sentence, the reply text, the first language, and the second language.