User-Specified Pronunciation Handling in Speech Synthesis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Digital assistants face challenges in accurately recognizing and synthesizing words with multiple pronunciations, particularly proper names, due to ambiguity in correct pronunciation, leading to recognition and synthesis errors.
Innovation Solution
Allowing users to specify preferred pronunciations for words, generating phonetic representations for both speech recognition and synthesis based on a single spoken input, and detecting errors to engage users in correcting pronunciations, enabling accurate association and storage of user-specified pronunciations for improved recognition and synthesis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If speech recognition and speech synthesis use different phonetic alphabets, then each process can be optimized for its specific function, but the system complexity increases and requires separate phonetic representations for the same word
Solution Approach 1:
The patent introduces a phonetic alphabet conversion module as an intermediary component that converts phonetic representations between speech recognition alphabet and speech synthesis alphabet. This mediator handles the complexity of maintaining separate phonetic systems by providing a standardized conversion interface, thus resolving the contradiction between reliability (optimized separate systems) and device complexity (multiple phonetic representations).
2Measurement precision
If users must provide multiple spoken inputs for training, then the system can gather more pronunciation data, but the training time and number of interactions increase
Solution Approach 1:
The patent applies preliminary action by using a single spoken input to pre-generate both the speech recognition phonetic representation and the speech synthesis phonetic representation. This preliminary processing of the input allows the system to create complete training data pairs in advance, eliminating the need for multiple separate recording sessions and reducing training time while maintaining pronunciation accuracy.
3Adaptability or versatility
If the digital assistant uses standard pronunciations, then synthesis is consistent, but it cannot accommodate user-specific or dialectal variations in pronunciation
Solution Approach 1:
The patent implements dynamics by making the phonetic representation system adaptable and changeable based on user input. The system dynamically generates or updates speech synthesis phonetic representations based on user-specified pronunciations, allowing it to transition from static standard pronunciations to dynamic user-customized pronunciations. This resolves the contradiction between adaptability (accommodating variations) and reliability (maintaining consistency) by providing both standard and customized pronunciation options.
Data Source
AI summary
The method is performed at an electronic device with one or more processors and memory storing one or more programs for execution by the one or more processors. A first speech input including at least one word is received. A first phonetic representation of the at least one word is determined, the first phonetic representation comprising a first set of phonemes selected from a speech recognition phonetic alphabet. The first set of phonemes is mapped to a second set of phonemes to generate a second phonetic representation, where the second set of phonemes is selected from a speech synthesis phonetic alphabet. The second phonetic representation is stored in association with a text string corresponding to the at least one word.