Speech Routing System Using Synthetic Feedback for New Skills

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing speech processing systems face challenges in dynamically routing user requests to the most appropriate skills, especially for new skills that lack sufficient user feedback, leading to suboptimal user experiences and inefficient learning of routing patterns.

Innovation Solution

A speech processing routing system that employs machine learning models to dynamically determine the appropriate action and speech processing components by using user feedback, contextual data, and predictive signals, including an exploration policy to route a portion of requests to new skills and update ranking models based on predicted user feedback.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If traditional routing methods are used that rely on existing user feedback, then routing accuracy for established skills is maintained, but new skills cannot be effectively routed due to lack of feedback data

Engineering Contradiction:

Improverouting capability for new skillsVSAvoidrouting accuracy

Core Design Contradiction:

Adaptability or versatilityVSMeasurement precision

Solution Approach 1:

The system performs preliminary actions by generating synthetic feedback data for new skills before actual user feedback is available. The feedback generation module creates predicted user feedback using machine learning models, allowing the routing system to learn and adapt to new skills proactively rather than waiting for accumulated real user feedback.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system introduces a feedback generation mechanism that creates synthetic feedback data for new skills. This feedback loop allows the routing model to continuously learn from both real user feedback and generated feedback, improving routing accuracy for new skills while maintaining performance for established skills.

Inventive Principle:

Principle #23Feedback

2Measurement precision

If user feedback is collected explicitly to improve routing accuracy, then routing performance improves, but user burden increases and user experience deteriorates

Engineering Contradiction:

Improverouting accuracyVSAvoiduser experience

Core Design Contradiction:

Measurement precisionVSEase of operation

Solution Approach 1:

The system implements self-service by automatically generating feedback data through machine learning models without requiring explicit user input. The feedback generation module autonomously creates predicted feedback signals based on user interactions, skill performance data, and contextual information, eliminating the need for users to manually provide feedback while still enabling continuous routing improvement.

Inventive Principle:

Principle #25Self-service

3Measurement precision

If more user feedback is collected to train routing models, then routing accuracy improves, but system complexity and data processing requirements increase

Engineering Contradiction:

Improverouting accuracyVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system introduces a feedback generation module as an intermediary that bridges the gap between limited real user feedback and the need for comprehensive training data. This module uses machine learning models to generate synthetic feedback data, reducing the dependency on collecting extensive real user feedback while maintaining routing model accuracy.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Reliability

If the system waits for sufficient user feedback before routing to new skills, then routing accuracy is maintained, but the ability to adapt and learn new skills is delayed

Engineering Contradiction:

Improverouting reliabilityVSAvoidskill learning speed

Core Design Contradiction:

ReliabilityVSAdaptability or versatility

Solution Approach 1:

The system performs preliminary learning actions by generating synthetic feedback data for new skills immediately upon their introduction. This allows the routing model to start learning about new skills without waiting for real user feedback to accumulate, accelerating the adaptation process while maintaining routing reliability through the use of predicted feedback signals.

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system implements a continuous feedback mechanism that combines real user feedback with generated feedback data. This dual-feedback approach allows the system to rapidly adapt to new skills by learning from generated feedback while progressively incorporating real user feedback to refine routing accuracy and maintain reliability.

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11289075B1Routing of natural language inputs to speech processing applications

Publication Date: 2022.03.29 AMAZON TECH INC

AI summary

Devices and techniques are generally described for using user feedback to determine routing decisions in a speech processing system. In various examples, first data representing a first utterance may be received. Second data representing a first semantic interpretation of the first utterance may be determined. A first intent data processing application may be selected for processing the second data. Feedback data may be determined related to the first intent data processing application processing the second data. Third data representing a semantic interpretation of a second utterance may be received, wherein the first semantic interpretation is the same as the second semantic interpretation. A second intent data processing application may be determined for processing the third data based at least in part on the feedback data.