Speech Recognition Disambiguation via Semantic Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Speech recognition systems face challenges in accurately processing spoken utterances due to ambiguity, leading to user frustration when multiple hypotheses are presented with no functional difference, as they fail to recognize semantic interpretations and spelling variations of entities.
Innovation Solution
A two-stage configuration using machine learning models to determine if disambiguation is necessary and which hypotheses to present to the user, with the first model deciding on execution and the second model selecting hypotheses for display, and pre-processing to remove semantically similar results before disambiguation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If speech recognition systems present multiple hypotheses to users for disambiguation, then the system provides more options for user selection, but the user experience deteriorates when the hypotheses are semantically similar or have no functional difference
Solution Approach 1:
The system performs preliminary semantic interpretation and spelling variation analysis before presenting hypotheses to users. By pre-processing the hypotheses to identify and remove semantically similar options, the system prepares only distinct, meaningful choices for user selection, thereby improving user experience while maintaining adaptability.
2Reliability
If the system presents all recognized hypotheses to the user, then completeness of options is improved, but the processing time and complexity increase due to redundant options
Solution Approach 1:
The system extracts and removes redundant hypotheses from the complete set of recognized options. By identifying and eliminating semantically similar hypotheses and spelling variations, the system retains only distinct options, thereby reducing processing time and complexity while maintaining the reliability of presenting complete meaningful options.
3Measurement precision
If the system performs semantic interpretation and removes similar results, then the accuracy of disambiguation is improved, but the device complexity increases due to additional processing stages
Solution Approach 1:
The system introduces an intermediary processing stage that performs semantic interpretation and similarity analysis between the speech recognition output and the disambiguation interface. This intermediary layer filters and prepares hypotheses by removing semantically similar options, thereby improving disambiguation accuracy while managing device complexity through a dedicated intermediate processing module.
Data Source
AI summary
Automatic speech recognition (ASR) processing including a feedback configuration to allow for improved disambiguation between ASR hypotheses. After ASR processing of an incoming utterance where the ASR outputs an N-best list including multiple hypotheses, the multiple hypotheses are passed downstream for further processing. The downstream further processing may include natural language understanding (NLU) or other processing to determine a command result for each hypothesis. The command results are compared to determine if any hypotheses of the N-best list would yield similar command results. If so, the hypothesis(es) with similar results are removed from the N-best list so that only one hypothesis of the similar results remains in the N-best list. The remaining non-similar hypotheses are sent for disambiguation, or, if only one hypothesis remains, it is sent for execution.