Proper Name Recognition Using Adaptive Grammar-Based ASR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic speech recognition (ASR) systems struggle to accurately recognize and understand spoken commands that include proper names, especially uncommon ones, often leading to incorrect or comical results.
Innovation Solution
A method and system that utilize a secondary grammar-based ASR system capable of rapid adaptation to recognize and understand proper names by generating an adaptation grammar for the secondary recognizer, which processes specific acoustic spans of the audio signal.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a fixed-vocabulary ASR system is used, then recognition accuracy for common words is high, but recognition accuracy for proper names deteriorates
Solution Approach 1:
The system segments the utterance into acoustic spans and identifies potential proper name entities within them. By separating proper name recognition from general vocabulary recognition, the system can apply specialized processing (secondary ASR with adaptation grammar) only to the relevant segments, thereby improving proper name recognition without compromising overall system efficiency.
Solution Approach 2:
The patent introduces an intermediary component (proper name adaptation grammar) that bridges the gap between the fixed-vocabulary primary ASR and the need for proper name recognition. This adaptation grammar serves as a mediator that enables the secondary ASR to recognize proper names by providing a customized vocabulary and structural framework.
2Measurement precision
If a secondary grammar-based ASR system with adaptation grammar is used, then proper name recognition accuracy is improved, but computational effort increases
Solution Approach 1:
The system applies different processing qualities to different parts of the input: the primary ASR handles general vocabulary recognition with standard processing, while the secondary ASR with adaptation grammar is applied locally only to acoustic spans containing proper names. This localized application of enhanced processing improves proper name recognition accuracy while minimizing additional computational effort.
Solution Approach 2:
Instead of applying full adaptation grammar processing to the entire utterance, the system performs partial action by applying it only to specific acoustic spans that contain proper names. This selective approach reduces computational overhead while still achieving high recognition accuracy for the critical proper name portions.
3Measurement precision
If adaptation is performed on the primary open dictation ASR system, then proper name recognition is improved, but system complexity and adaptation time increase
Solution Approach 1:
The patent extracts the adaptation functionality from the primary open dictation ASR system and places it in a separate secondary ASR system. By taking out the adaptation grammar and processing logic from the primary system, the invention avoids increasing the complexity of the well-established primary ASR while still enabling proper name recognition through the secondary system.
Solution Approach 2:
The secondary ASR system with adaptation grammar serves multiple functions: it recognizes proper names, handles uncommon vocabulary, and provides a framework for customized recognition without requiring modifications to the primary ASR system. This multi-functionality approach allows a single added component to address various recognition challenges.
4Measurement precision
If extensive adaptation is performed, then recognition accuracy for proper names is improved, but processing latency increases
Solution Approach 1:
The system performs preliminary action by preparing the adaptation grammar in advance and having the secondary ASR system ready to process acoustic spans. The adaptation grammar is pre-compiled and structured, allowing rapid processing when proper names are detected, thereby minimizing additional latency while maintaining high recognition accuracy.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Various embodiments contemplate systems and methods for performing automatic speech recognition (ASR) and natural language understanding (NLU) that enable high accuracy recognition and understanding of freely spoken utterances which may contain proper names and similar entities. The proper name entities may contain or be comprised wholly of words that are not present in the vocabularies of these systems as normally constituted. Recognition of the other words in the utterances in question, e.g. words that are not part of the proper name entities, may occur at regular, high recognition accuracy. Various embodiments provide as output not only accurately transcribed running text of the complete utterance, but also a symbolic representation of the meaning of the input, including appropriate symbolic representations of proper name entities, adequate to allow a computer system to respond appropriately to the spoken request without further analysis of the user's input.