Pronunciation Feedback via Phoneme Sequence Comparison
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional language learning software cannot provide precise feedback on pronunciation errors, only indicating that a user's pronunciation is incorrect without specifying the exact issue, hindering effective learning.
Innovation Solution
A method and device that transform input voice into phoneme sequences, compare them using dynamic time warping (DTW) algorithms with standard phoneme sequences, and generate ancillary information to identify and highlight specific pronunciation errors, enabling users to understand and correct their mistakes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If conventional pronunciation comparison is used, then the system can determine whether pronunciation is correct, but the user cannot understand the specific pronunciation errors
Solution Approach 1:
The patent segments the pronunciation comparison process into phoneme-level analysis. Instead of treating the entire word or sentence as a single unit, the system breaks it down into individual phonemes, compares each phoneme separately, and identifies specific phoneme-level errors. This segmentation allows the system to provide detailed feedback on which specific phonemes are mispronounced, resolving the contradiction between providing comprehensive error information and maintaining measurement precision.
Solution Approach 2:
The patent introduces phoneme sequences as an intermediary representation between the raw audio input and the final pronunciation judgment. The audio signal is first converted into phoneme sequences, which then serve as the basis for comparison with standard pronunciations. This intermediary layer enables precise identification of specific phoneme errors while maintaining overall pronunciation assessment accuracy.
2Loss of information
If detailed phoneme-level analysis is implemented, then specific pronunciation errors can be identified, but the system complexity increases
Solution Approach 1:
The patent uses phoneme sequences as simplified copies or representations of the actual audio signals. Instead of performing complex direct audio signal processing and comparison, the system converts audio into phoneme sequences (which are easier to handle and compare) and performs the comparison operation on these copied representations. This copying approach provides detailed pronunciation feedback while avoiding the need for overly complex audio processing systems.
Data Source
AI summary
A method and a device for learning a language and a computer readable recording medium are provided. The method includes following steps. An input voice from a voice receiver is transformed into an input sentence according to a grammar rule. Whether the input sentence is the same as a learning sentence displayed on a display is determined. If the input sentence is different from the learning sentence, an ancillary information containing at least one error word in the input sentence that is different from the learning sentence is generated.