Shared Backbone Neural Network for Competency Evaluation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional competency evaluation methods are time-consuming and costly, requiring specialized training and expertise, and lack objectivity, making it necessary to develop an efficient method for evaluating competencies using a machine learning model.
Innovation Solution
A method involving a machine learning model with a backbone artificial neural network module for deriving intermediate feature information and a sub-artificial neural network module for evaluating competencies, which includes a labeling learning step to reduce errors between prediction information and labeling information, allowing for automatic evaluation of competencies from input data such as text, video, or voice.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional competency evaluation methods are used with specialized evaluators, then evaluation accuracy and objectivity are improved, but evaluation time and cost increase significantly
Solution Approach 1:
The patent replaces the mechanical system of human evaluators conducting detailed assessments with an automated machine learning evaluation system. The ML model processes evaluation data through neural network modules (backbone module for feature extraction, evaluation module for competency assessment) to automatically generate competency evaluations, eliminating the need for specialized human evaluators and significantly reducing evaluation time while maintaining objectivity
Solution Approach 2:
The patent creates a computational copy of the evaluation process by training the machine learning model on labeled evaluation data. The model learns to replicate the decision-making patterns of expert evaluators through supervised learning, allowing it to produce evaluations that mirror expert assessments without requiring actual expert involvement in each evaluation case
2Measurement precision
If conventional competency evaluation methods are used with specialized evaluators, then evaluation quality is improved, but cost increases due to training and hiring experts
Solution Approach 1:
The patent replaces the expensive human expert system with an automated machine learning system. Once the ML model is trained on labeled data from expert evaluations, it can perform evaluations at minimal marginal cost without requiring ongoing expert training or hiring, significantly reducing the energy and financial resources required for each evaluation
Solution Approach 2:
The patent performs preliminary training of the machine learning model using labeled evaluation data before deployment. This upfront investment in training the backbone neural network module and evaluation neural network module with comprehensive labeled datasets enables the system to achieve expert-level evaluation quality without requiring continuous expert involvement, reducing long-term operational costs
3Measurement precision
If a machine learning model with separate networks for each competency is used, then evaluation accuracy for each competency is improved, but computational load and training time increase
Solution Approach 1:
The patent merges multiple competency evaluation functions into a unified machine learning architecture. The backbone artificial neural network module serves as a shared feature extraction engine that processes input data once and generates intermediate features that are then used by the evaluation neural network module for multiple competency assessments simultaneously, reducing redundant computation and training time
Solution Approach 2:
The backbone neural network module is designed with universal functionality to extract features applicable across multiple competencies. This single multi-functional module replaces what would otherwise require separate feature extraction networks for each competency, reducing the overall computational burden and training requirements while maintaining accurate evaluation capabilities for each specific competency
Data Source
AI summary
The present invention relates to a method, a device and a computer-readable medium for training a machine learning model performing competency evaluation on a plurality of competencies, and more particularly, to a method, a device and a computer-readable medium for training a machine learning model performing competency evaluation on a plurality of competencies to efficiently train the machine learning model for outputting an evaluation result of each of the competencies from input data related to answers and the like of an evaluatee.