Natural Language Classifier Training via User Feedback

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current mobile applications for improving emotional and physical well-being rely heavily on programming and require human administrators for diagnostics and optimization, limiting their accuracy and efficiency in providing interactive and personalized user experiences.

Innovation Solution

A computing system with a user-interface and natural language classifier that accesses a database of dialogues, retrieves user responses with confidence scores, allows user validation of assigned classes, updates responses based on user feedback, and stores new data to re-train the classifier, enhancing the system's emotional intelligence and interaction quality.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If mobile applications employ natural language processing to understand user reactions, then the accuracy of understanding user responses is improved, but the complexity of programming and system operation increases

Engineering Contradiction:

Improveaccuracy of understanding user responsesVSAvoidcomplexity of programming and system operation

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The system implements a feedback mechanism where administrators can review classified user responses, provide corrections, and the system learns from these corrections to improve its natural language processing accuracy over time

Inventive Principle:

Principle #23Feedback

Solution Approach 2:

The system introduces an intermediary layer of automated classification that handles the complex natural language processing, allowing administrators to operate simpler interfaces while still benefiting from advanced NLP capabilities

Inventive Principle:

Principle #24Intermediary (Mediator)

2Reliability

If human administrators review and optimize mobile applications, then the reliability of the system is improved, but the loss of time and productivity decreases

Engineering Contradiction:

Improvereliability of system performanceVSAvoidtime required for diagnostics and optimization

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The system performs self-optimization by automatically reviewing its own performance through the administered interface, allowing it to diagnose and optimize its natural language processing without requiring extensive manual intervention

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The system uses feedback from administrator reviews to automatically adjust and optimize its classification accuracy, reducing the need for continuous manual tuning and optimization

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS11779270B2Systems and methods for training artificially-intelligent classifier

Publication Date: 2023.10.10 TWILL INC

AI summary

A computing system/method for enabling a user to improve, via training, a system designed to increase the emotional and/or physical well-being of persons, or designed for other purposes. The system/method includes retrieving a user response from a dialogue database, the user response having already labeled thereto an assigned class having a highest confidence score, the confidence score indicating degree of confidence that context of the retrieved user response is of the assigned class, displaying, the assigned class, along with other classes each having a respective lower confidence score, and receiving an indication of validity of the assigned class. The system/method further includes retrieving of a pair of sequential user response and follow-up prompt from the database, displaying user-selectable ratings, each rating designating a respectively different quality to the follow-up prompt, receiving selection of a rating and a related comment, and associating the selection and the comment to the follow-up prompt.