Adaptive Pronunciation Dictionary for Speech Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech recognition systems face difficulties in accurately recognizing voice commands when users pronounce words differently from the canonical pronunciations stored in pronunciation dictionaries, leading to reduced recognition accuracy.
Innovation Solution
A personalized pronunciation dictionary that adapts to a user's unique pronunciation by learning from their interactions, allowing the system to associate non-canonical pronunciations with entity names, thereby improving speech recognition and transcription quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a canonical pronunciation dictionary is used for speech recognition, then the system has a standardized pronunciation reference, but it cannot accurately recognize user-specific non-canonical pronunciations
Solution Approach 1:
The pronunciation dictionary is transformed from a static canonical reference to a dynamic adaptive structure that evolves based on user interactions. The system continuously updates pronunciations associated with entity names by analyzing user corrections and subsequent confirmations, allowing the dictionary to adapt to individual user pronunciation patterns while maintaining its core organizational structure.
Solution Approach 2:
The system implements a feedback mechanism where user corrections to speech recognition results are captured and analyzed. When a user corrects a misrecognized entity name, the system uses this correction as feedback to learn the user's intended pronunciation. Subsequent unrecognized utterances are compared against this learned pronunciation, and confirmed matches trigger dictionary updates, creating a closed-loop feedback system that continuously improves recognition accuracy.
2Measurement precision
If the pronunciation dictionary is updated with user-specific pronunciations, then recognition accuracy for that user improves, but the system complexity increases
Solution Approach 1:
The system applies local quality by maintaining different pronunciation characteristics for different users and entity names. Each entity name can have multiple pronunciations stored in the dictionary, with each pronunciation associated with specific user identifiers. This allows the dictionary to have localized user-specific pronunciations while maintaining the global canonical pronunciation structure, enabling targeted adaptation without complete system restructuring.
Solution Approach 2:
The pronunciation dictionary is segmented into multiple independent pronunciation entries for each entity name, rather than maintaining a single canonical pronunciation. Each pronunciation variant is stored as a separate record with associated metadata including user identifiers and confidence scores. This segmentation allows the system to manage multiple pronunciations efficiently and select the appropriate one based on user context without increasing overall system complexity.
Data Source
Figure 1A~1B
Figure 2
Figure 3
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer-storage medium, for implementing a pronunciation dictionary. The method includes receiving audio data corresponding to an utterance that includes a command and an entity name. Additionally, the method may include generating, by an automated speech recognizer, an initial transcription for a portion of the audio data that is associated with the entity name, receiving a corrected transcription for the portion of the utterance that is associated with the entity name, obtaining a phonetic pronunciation that is associated with the portion of the audio data that is associated with the entity name, updating a pronunciation dictionary to associate the phonetic pronunciation with the entity name, receiving a subsequent utterance that includes the entity name, and transcribing the subsequent utterance based at least on the updated pronunciation dictionary. Improved speech recognition and more higher quality transcription can be provided.