Speech Recognition Adaptation via Utterance Concatenation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Short utterances in speech recognition systems often result in low accuracy due to mismatches between acoustic and language models, and existing systems require multiple sessions or speaker identification for adaptation, which affects scalability.
Innovation Solution
A method that concatenates short utterances with additional utterances to enhance recognition accuracy by transmitting the concatenated utterance to a speech recognition server, allowing for improved transcription without the need for speaker identification or multiple sessions, using client software to extract the original utterance's transcription based on time stamps or text association.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If the system uses short utterances for speech recognition, then the user interaction speed is improved, but the speech recognition accuracy deteriorates
Solution Approach 1:
The patent combines multiple short utterances into a single concatenated utterance for processing. By merging several short speech segments together, the system maintains the speed advantage of short utterances while achieving the accuracy benefits of longer speech samples through the combined phonetic and contextual information.
Solution Approach 2:
The system performs preliminary concatenation of short utterances before submitting to the speech recognition server. This preliminary action of combining utterances client-side allows the system to prepare enhanced input data without requiring multiple separate recognition passes, thus maintaining interaction speed while improving accuracy.
2Measurement precision
If the system performs adaptation across multiple user sessions, then the speech recognition accuracy is improved, but the storage requirements and system complexity increase
Solution Approach 1:
The patent segments the adaptation process into client-side concatenation operations rather than server-side multi-session processing. By dividing the task of improving accuracy into local utterance combining operations, the system avoids the complexity of managing multiple sessions and user identification while still achieving adaptation benefits.
Solution Approach 2:
The client software acts as an intermediary that performs concatenation of utterances before transmission to the speech recognition server. This intermediary function eliminates the need for complex server-side session management and user identification mechanisms, simplifying the overall system while maintaining accuracy improvement capabilities.
3Measurement precision
If the system stores each user's utterances for adaptation, then the speech recognition accuracy is improved, but the storage space requirements increase
Solution Approach 1:
The system performs partial adaptation by concatenating only the necessary utterances locally rather than storing and processing all user utterances. This partial action approach achieves sufficient accuracy improvement for the given task without the excessive storage requirements of comprehensive utterance archiving.
Data Source
AI summary
A method, computer program product, and system for adapting speech recognition of a user's speech is provided. The method includes receiving a first utterance from a user having a duration below a predetermined threshold, identifying at least one further utterance from the user that provides additional information, generating a concatenated utterance by concatenating the first utterance with the at least one further utterance, transmitting the concatenated utterance to a speech recognition server, receiving a transcription of the concatenated utterance from the speech recognition server that includes a transcription of the first utterance, and extracting the transcription of the first utterance from the transcription of the concatenated utterance. The transcription of the first utterance is based on the additional information provided by the at least one further utterance.