Semi-Supervised Learning System With Preliminary Attribute Verification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In semi-supervised learning, it is challenging to set a threshold value that effectively eliminates symbol information not originally an attribute value, leading to insufficient enhancement of learner accuracy.
Innovation Solution
A learning system that includes means to classify symbol information, acquire attribute value candidates, determine whether they satisfy predetermined conditions, and control additional learning based on these determinations to prevent non-attribute values from being added to training data, using methods such as checking character count, specific symbol types, and appearance frequency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If symbol information with probability equal to or more than threshold value is added to training data, then training data is expanded, but non-attribute values may be erroneously added reducing accuracy
Solution Approach 1:
The determination unit performs preliminary verification of attribute value candidates before they are added to training data. It checks whether candidates satisfy predetermined conditions (such as appearing in specific document sections, having appropriate length, or matching expected formats) to filter out non-attribute values before they contaminate the training dataset.
Solution Approach 2:
The determination unit acts as an intermediary between the attribute value acquisition unit and the training data. It mediates the addition process by verifying candidates against predetermined conditions, serving as a gatekeeper that allows only valid attribute values to be added while blocking non-attribute values.
2Manufacturing precision
If threshold value is set to eliminate non-attribute values, then accuracy may be improved, but it is very difficult to set an appropriate threshold value
Solution Approach 1:
Instead of relying on a single probability threshold, the system changes the verification parameters by checking multiple predetermined conditions (such as document section location, character length range, format patterns, or frequency of occurrence). This multi-parameter approach replaces the difficult-to-set single threshold with several easier-to-determine conditions.
Solution Approach 2:
The verification process is segmented into multiple independent checks rather than relying on a single threshold comparison. Each predetermined condition represents a separate verification dimension (e.g., positional check, length check, format check), making the overall system more manageable and easier to configure.
3Manufacturing precision
If manual preparation of highly accurate training data is performed, then training data quality is improved, but it takes a lot of time and effort
Solution Approach 1:
The learner performs self-service by automatically acquiring attribute value candidates from documents and verifying them against predetermined conditions. This automated self-verification process eliminates the need for manual preparation of high-quality training data, allowing the system to generate accurate training data autonomously without human intervention.
Solution Approach 2:
The system performs preliminary automatic verification of attribute value candidates before they are added to training data. This preliminary filtering action ensures data quality without requiring manual preparation, as the predetermined conditions automatically eliminate non-attribute values during the acquisition phase.
Data Source
AI summary
To enhance the accuracy of a learner in semi-supervised learning, learning means of a learning system (S) causes the learner, which is configured to classify symbol information included in each of a plurality of documents, to learn based on training data indicating an attribute value of each of a plurality of attributes. Acquisition means inputs each of the plurality of documents to the learner to acquire the symbol information classified by the learner as an attribute value candidate. Determination means determines whether a symbol or a symbol string indicated by the attribute value candidate satisfies a predetermined condition. Additional learning control means controls, based on a determination result obtained by the determination means, additional learning by the learner using the attribute value candidate.