Automatic Speech Recognition Sentence Boundary Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current speech recognition systems face challenges in accurately determining sentence boundaries, especially in naturally spoken sentences without silence, and incorrectly identifying pauses between sentences.
Innovation Solution
An automatic speech recognition device that captures speech data, recognizes phonemes, performs syntactic parsing, and uses a combination of acoustic, word statistics, and syntactic coherence scores to determine sentence boundaries through modules like phoneme recognition, word recognition, and end-of-sentence determination, employing techniques such as deep neural networks and support vector machines.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sentence boundaries are determined based on pause or silence detection, then the system is simple to implement, but it produces false boundaries in continuous speech without silence and misses boundaries where silence is absent
Solution Approach 1:
The patent combines multiple detection approaches (acoustic pause detection, word statistics analysis, and syntactic coherence checking) into a unified sentence boundary detection system. This merging allows the system to overcome the limitations of individual methods by using their complementary strengths, thereby improving detection accuracy without requiring a completely new complex system
Solution Approach 2:
The patent introduces syntactic coherence as an intermediary layer between acoustic detection and final boundary determination. This intermediary analyzes the grammatical structure and semantic relationships of words to validate or correct boundaries suggested by acoustic cues, improving accuracy while maintaining a modular system architecture
2Measurement precision
If multiple scoring methods (acoustic, word statistics, syntactic coherence) are combined to determine sentence boundaries, then detection accuracy improves, but computational complexity increases
Solution Approach 1:
The patent segments the sentence boundary detection task into three independent scoring components: acoustic scoring, word statistics scoring, and syntactic coherence scoring. Each component processes different aspects of the speech signal independently, allowing for parallel computation and reducing the complexity of any single processing module while maintaining overall detection accuracy
Solution Approach 2:
The patent transforms the complex multi-criteria decision problem into a parameter-based scoring system where each detection criterion is converted into a numerical score. By changing the representation from qualitative linguistic rules to quantitative parameters, the system enables efficient computational processing through weighted combination of scores
Data Source
AI summary
Technologies for detecting an end of a sentence in automatic speech recognition are disclosed. An automatic speech recognition device may acquire speech data, and identify phonemes and words of the speech data. The automatic speech recognition device may perform a syntactic parse based on the recognized words, and determine an end of a sentence based on the syntactic parse. For example, if the syntactic parse indicates that a certain set of consecutive recognized words form a syntactically complete and correct sentence, the automatic speech recognition device may determine that there is an end of a sentence at the end of that set of words.