Unseen Class Score Calculation for Classification Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing classification systems struggle to accurately classify data into unseen classes for which no labeled data exist, as they tend to further reduce the already low probability of classification into such classes.
Innovation Solution
A classification system that includes input means for document data, a hierarchical structure, unseen class score calculation means, matching score calculation means, and final classification score calculation means, which calculates a final classification score by combining the classification scores of seen and unseen classes with matching scores, thereby improving the classification accuracy of unseen classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a general hierarchical classifier is used to classify data, then classification accuracy for seen classes is maintained, but classification accuracy for unseen classes deteriorates because the probability of classification into unseen classes is already low and is further reduced
Solution Approach 1:
The classification process is segmented into two independent scoring components: (1) classification score from the hierarchical classifier that evaluates seen classes, and (2) matching score from semantic similarity analysis that evaluates both seen and unseen classes. These segmented scoring mechanisms are then combined to produce the final classification result, allowing each component to specialize and thereby improving overall performance on unseen classes without sacrificing seen class accuracy.
Solution Approach 2:
A matching score calculation mechanism is introduced as an intermediary component that bridges the gap between the hierarchical classifier and unseen class classification. This intermediary calculates semantic similarity between test data and class labels using embedding vectors, providing an additional evaluation pathway that compensates for the hierarchical classifier's inability to recognize unseen classes, thereby improving adaptability while maintaining reliability.
2Reliability
If a process is performed to lower the probability of unseen class classification further, then the probability distribution becomes more confident for seen classes, but the ability to correctly classify unseen classes is lost
Solution Approach 1:
The final classification score is obtained by merging two independent scoring mechanisms: the classification score from the hierarchical classifier (which provides reliability for seen classes) and the matching score from semantic similarity analysis (which provides accuracy for unseen classes). This combination allows the system to maintain confidence in seen class classification while simultaneously improving the ability to correctly classify unseen classes, resolving the trade-off between reliability and measurement precision.
Data Source
AI summary
The input means 181 accepts inputs of test data, a hierarchical structure in which a node of bottom layer represents a target class, and a classification score of a seen class as the classification score indicating a probability that the test data is classified into each class. The unseen class score calculation means 182 calculates the classification score of an unseen class based on uniformity of the classification score of each seen class. The matching score calculation means 183 calculates a matching score indicating similarity between the test data and each class label. The final classification score calculation means 184 calculates a final classification score indicating a probability that the test data is classified into the class so that the larger the classification score of each class, and the matching score, the larger the final classification score.