NLU Model Generation via Data Expansion

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current speech processing systems lack an efficient developer experience for generating and managing Natural Language Understanding (NLU) models, particularly in terms of authoring, testing, and releasing data structures for runtime operations across multiple skills and devices, which hinders the development of robust and language-agnostic NLU models.

Innovation Solution

A unified NLU modeling system with a web-based developer interface allows developers to input, validate, and release data structures for NLU models, expanding developer-provided data to create more robust models by generating alternate phrasings and translations, and facilitating transitions between different NLU model generation systems.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If developers manually create and manage NLU model data structures for each skill and device, then the system can support multiple skills and devices, but the development time and complexity increase significantly

Engineering Contradiction:

Improvesupport for multiple skills and devicesVSAvoiddevelopment time

Core Design Contradiction:

Adaptability or versatilityVSLoss of time

Solution Approach 1:

The patent implements a universal NLU model generation system that can serve multiple skills and devices through a single interface. The system processes developer inputs and generates data structures that are applicable across different skills and device types, eliminating the need to manually create separate models for each skill or device.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system performs preliminary expansion of developer-provided data by automatically generating alternate phrasings and translations before the models are deployed. This pre-processing step creates a more robust NLU model that is ready for multiple skills and devices without requiring additional manual work during the deployment phase.

Inventive Principle:

Principle #10Preliminary action

2Productivity

If developers provide minimal data for NLU model generation, then the development process is faster, but the model accuracy and robustness decrease

Engineering Contradiction:

Improvemodel generation speedVSAvoidmodel robustness

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system automatically expands and enriches the developer-provided data without requiring manual intervention. It generates alternate phrasings, translations, and variations autonomously, allowing the developer to maintain fast productivity while the system handles the robustness enhancement independently.

Inventive Principle:

Principle #25Self-service

3Adaptability or versatility

If the system generates comprehensive data structures with alternate phrasings and translations, then language-agnostic NLU models are achieved, but the processing complexity increases

Engineering Contradiction:

Improvelanguage-agnostic capabilityVSAvoidprocessing complexity

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent segments the NLU model generation process into distinct modular components: receiving developer input, validating the input, expanding to generate alternate phrasings and translations, and releasing the final data structures. This modular architecture manages processing complexity by organizing tasks into separate, manageable stages.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system introduces an intermediary validation step that checks developer inputs against a data scheme before processing. This intermediary layer ensures data quality and consistency, reducing the complexity of downstream processing by preventing invalid or inconsistent data from propagating through the system.

Inventive Principle:

Principle #24Intermediary (Mediator)

4Manufacturing precision

If developers validate data structures against strict data schemes, then model quality is ensured, but the validation time increases

Engineering Contradiction:

Improvedata structure qualityVSAvoidvalidation time

Core Design Contradiction:

Manufacturing precisionVSLoss of time

Solution Approach 1:

The system performs validation as a preliminary step before expanding the developer data. By validating early, it ensures that only correct data structures proceed to the expansion phase, preventing rework and reducing overall processing time while maintaining high data quality standards.

Inventive Principle:

Principle #10Preliminary action

Data Source

PatentUS11211056B1Natural language understanding model generation

Publication Date: 2021.12.28 AMAZON TECH INC

AI summary

Systems and techniques for generating natural language understanding (NLU) models are described. A developer of an NLU model may provide data representing runtime NLU functionality. For example, a developer may provide one or more sample natural language user inputs. The NLU model generation system may expand data, provided by the developer, to result in a more robust NLU model for use at runtime. For example, the NLU model generation system may expand sample natural language user inputs, may translate sample natural language user inputs into other languages, etc. The present disclosure also provides a mechanism for transitioning between using NLU models of a first NLU model generation system and NLU models of a second NLU model generation system.