Voice Recognition Politeness Detection Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current voice recognition systems face inaccuracies when dealing with unclear or soft-spoken inputs, particularly from children, and require additional training and complex architecture to handle polite expressions, leading to reduced accuracy and user frustration.
Innovation Solution
A natural language model-based voice recognition system that minimally trains and adapts for legacy systems, using a binary classification model to distinguish between queries with and without obsequious expressions, allowing for plug-and-play modularity and improved recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If voice recognition systems require polite expressions (please, thank you) to be included in commands, then device responsiveness to polite speech improves, but speech recognition accuracy deteriorates due to reduced match scores
Solution Approach 1:
The system segments the speech processing pipeline into distinct components: a politeness detection module that identifies polite expressions, and a speech recognition module that processes the core command. The polite expressions are separated and handled independently, allowing the speech recognition to focus on accurate command interpretation while the politeness detection manages social conventions.
Solution Approach 2:
An intermediary processing layer is introduced between speech input and command execution. This layer detects and manages polite expressions, determining whether they should be preserved, removed, or modified before the speech reaches the recognition engine. This intermediary prevents polite expressions from interfering with match scores while maintaining device responsiveness to polite speech.
2Reliability
If systems are trained to recognize queries with obsequious expressions, then polite speech recognition improves, but system complexity increases and legacy match scores are reduced
Solution Approach 1:
The system divides functionality into separate modules: a lightweight politeness detection component that identifies polite expressions using simple pattern matching or classification, and the existing speech recognition system that handles command processing. This segmentation avoids the need for complex retraining of the entire system while improving polite speech recognition.
Solution Approach 2:
Instead of completely retraining the speech recognition system to handle all variations of polite speech, the system applies a partial solution by adding a dedicated politeness detection layer that handles only the polite expression aspect. This partial action approach reduces system complexity compared to full retraining while achieving the desired reliability for polite speech.
3Device complexity
If obsequious expressions are removed from queries to simplify processing, then system architecture simplifies, but content search accuracy deteriorates for titles containing such expressions
Solution Approach 1:
The system applies different processing rules to different parts of the query based on local context. Polite expressions are identified and handled differently depending on whether they are part of the command structure or part of the content being searched. This local quality approach ensures that expressions like 'please' in 'play Barney, please' are managed appropriately while preserving accuracy for titles like 'Play Thank You for Smoking'.
Solution Approach 2:
An intermediary analysis step examines each detected polite expression to determine its function: whether it modifies the command intent or is part of the search term itself. This intermediary decision-making process preserves necessary expressions for accurate content matching while removing or modifying those that would interfere with command processing, maintaining both architectural simplicity and search accuracy.
Data Source
AI summary
Systems and methods for training a natural language model of a natural language understanding (NLU) system are disclosed herein. A text string including at least a content entity is received. A determination is made as to whether the text string includes an obsequious expression. In response to determining the text string includes an obsequious expression, a determination is made as to whether the obsequious expression describes the content entity. A query is forwarded in response to determining the text string includes an obsequious expression and in determining the obsequious expression describes the content entity. In response to determining the obsequious expression describes the content entity, the query includes the obsequious expression and in response to determining the obsequious expression does not describe the content entity, the query does not include the obsequious expression.