Decision Model Training via Rationale Vector Assembly
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing machine-learning methods for training decision-making models with natural language data face challenges in adapting to different domains due to the need for domain-specific questionnaires and lack of detailed information for distinguishing similar semantic topics, leading to inefficiencies and biases in training.
Innovation Solution
A machine-learning method that constructs effective vector groups from labeled natural language text files by connecting rationale vector groups and using supervised classification algorithms to train decision-making models, allowing for the labeling of unlabeled text files, and employing data augmentation techniques to generate additional training material.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If domain-specific questionnaires are used to manually label categorical data, then the decision-making model can be trained with structured information, but the method cannot be easily adapted to various domains and requires significant manual effort
Solution Approach 1:
The patent applies universality by designing a questionnaire generation system that can automatically adapt to multiple domains. The system uses pre-stored question templates and knowledge graphs that can be configured for different domains (e.g., court cases, medical records) without requiring complete redesign, enabling one system to serve multiple functions across various domains while maintaining labeling precision
Solution Approach 2:
The patent applies preliminary action by pre-storing question templates, answer options, and knowledge graph structures before actual labeling tasks. These pre-configured elements can be quickly retrieved and adapted for different domains, eliminating the need for manual questionnaire design for each new domain while preserving labeling accuracy
2Productivity
If semantic topic classification methods (LSA, LDA) are used to process natural language data, then longer text files can be classified into semantic topics, but more detailed information is needed to distinguish texts of similar semantic topics
Solution Approach 1:
The patent applies segmentation by breaking down the natural language processing into multiple stages: first extracting key entities and terms, then generating targeted questions based on these elements, and finally creating structured labels. This segmented approach preserves detailed information while maintaining processing efficiency by focusing computational resources on critical aspects of the text
Solution Approach 2:
The patent introduces an intermediary mechanism - the generated questionnaire - that bridges the gap between raw natural language text and structured labels. This intermediary preserves detailed information by capturing specific entities, relationships, and contextual elements that would otherwise be lost in direct classification, while still enabling efficient processing through automated question-answer pairing
3Reliability
If manual labeling of natural language data is performed to obtain categorical data, then the data quality for training is improved, but the training efficiency and time consumption are reduced
Solution Approach 1:
The patent applies self-service by enabling the system to automatically generate questionnaires and labels from raw natural language data without requiring manual annotators. The system uses pre-stored knowledge graphs and templates to self-generate training data, maintaining high data quality through structured generation while dramatically improving training efficiency and reducing time consumption
Data Source
AI summary
A machine-learning method for training a decision-making model includes: obtaining a rationale vector group for a rationale included in a labeled natural language text file; assembling an effective vector group for the labeled natural language text file by connecting the rationale vector groups for the rationales using a specific order; and executing a supervised classification algorithm to train the decision-making model using the effective vector group and a target decision for the natural language text file. The decision-making model is trained to be configured to label an unlabeled natural language text file using one of a plurality of potential decisions that serves as a target decision.