Utterance Generation and Evaluation for Speech Command Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing speech-controlled computing systems face challenges in efficiently processing user commands due to rigid invocation patterns, which limit user freedom and increase cognitive burden, as well as resource usage, when invoking specific applications.
Innovation Solution
An utterance generation and evaluation system that uses historical transaction data and machine-learned models to determine related intents and targets, generating alternative utterance phrases with higher confidence scores for effective skill invocation, thereby reducing the need for precise invocation patterns and improving user experience.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If rigid invocation patterns are used for speech commands, then the system can process commands efficiently with clear structure, but user freedom is limited and cognitive burden increases
Solution Approach 1:
The system dynamically adapts the invocation pattern recognition from rigid to flexible. Instead of requiring exact pattern matches, the machine-learned model learns from historical transaction data to recognize varied user utterances that convey the same intent, allowing the system to handle both structured and natural speech patterns effectively
Solution Approach 2:
The system changes the parameter of pattern matching from exact string matching to probabilistic intent recognition. By using machine-learned models trained on historical data, the system transforms the recognition parameter from deterministic pattern matching to flexible intent classification, enabling better user freedom while maintaining processing efficiency
2Ease of operation
If multiple alternative utterance phrases are generated and evaluated, then user interaction quality improves, but system resource usage increases
Solution Approach 1:
The system performs preliminary action by pre-processing historical transaction data to train machine-learned models before actual command processing. This upfront preparation enables the system to quickly evaluate alternative utterances during runtime without excessive resource consumption, as the heavy computational work is done in advance
Solution Approach 2:
The system uses its own historical transaction data to train and improve its utterance evaluation capabilities. By leveraging accumulated user interaction data, the system continuously self-improves its ability to generate and evaluate alternative utterances without requiring external resources or manual intervention
Data Source
AI summary
Devices and techniques are generally described for generating and evaluating utterances. In some examples, an utterance generation and evaluation system can receive intent data and target data. The utterance generation and evaluation system can determine related target names and related intent names and, based on the related target names and related intent names, can generate an utterance phrase. The utterance generation and evaluation system can determine a confidence score associated with the utterance phrase and, based on the confidence score, determine the utterance phrase as a recommended utterance phrase.