Speech Contact Identification via Collapsed String Matching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Voice recognition systems in mobile devices and in-vehicle infotainment systems face difficulties in accurately identifying unique names and words, leading to misinterpretation of voice commands.
Innovation Solution
A method that involves a speech recognition module to generate candidate strings with confidence scores, a logic module to evaluate and correct these strings by comparing them to user contacts, and a process of collapsing strings for simplified matching, along with filtering based on textual similarity, location, and time to improve contact identification in voice commands.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If voice recognition systems use standard recognition algorithms, then the system is simple to implement, but the accuracy of identifying unique names and words deteriorates
Solution Approach 1:
The system segments the contact identification process into multiple stages: initial voice recognition to generate candidate strings, confidence score evaluation, string collapsing for simplified matching, and metadata-based filtering. This segmentation allows each stage to focus on specific tasks, improving overall accuracy without requiring a completely complex system redesign.
Solution Approach 2:
The system introduces intermediary components including confidence scores as mediators between voice recognition and contact matching, string collapsing as an intermediary representation layer, and metadata (location, time, textual similarity) as intermediary filters. These intermediaries bridge the gap between simple recognition and accurate identification.
2Measurement precision
If the system performs detailed string comparison and filtering, then contact identification accuracy improves, but processing time increases
Solution Approach 1:
The system performs preliminary actions by generating confidence scores for candidate strings and collapsing them into simplified forms before detailed comparison. Metadata such as location, time, and textual similarity are pre-computed and stored, allowing rapid filtering during the actual recognition process without time-consuming real-time analysis.
Solution Approach 2:
The system applies different levels of analysis to different aspects of contact identification: quick confidence score evaluation for overall match quality, simplified string collapsing for structural comparison, and targeted metadata filtering for contextual relevance. This localized quality approach ensures thorough accuracy where needed while maintaining efficiency in other areas.
3Reliability
If the system uses multiple filtering criteria (textual similarity, location, time), then identification reliability improves, but system complexity increases
Solution Approach 1:
The system achieves multi-functionality by using a unified metadata structure that can simultaneously store and evaluate multiple criteria (textual similarity, location, time). This universal approach allows the same filtering framework to handle different types of information without requiring separate complex systems for each criterion.
Solution Approach 2:
The system changes parameters by transforming voice recognition output into confidence scores, converting contact information into structured metadata with specific parameters (location coordinates, time stamps, similarity metrics). These parameter transformations enable systematic filtering while maintaining reliability through consistent evaluation criteria.
Data Source
AI summary
A system receives candidate strings from a speech recognition engine. Where the speech recognition indicates success, the candidate string may be reported or otherwise used. Otherwise, candidate strings may be processed by comparing collapsed forms of the candidate strings to collapsed forms of a user's contacts. The collapsed form may replace contiguous consonants with a 1 and contiguous vowels with a 0. If a candidate string has a collapsed form matching a contact's collapsed form, that contact may be used. Where multiple candidate strings match the collapsed forms of multiple contacts, the contacts may be filtered according to textual similarity to the candidate strings, proximity to a typical usage location for the contacts, and a typical usage time for the contacts.