Prompt Classifier Cascades for Imbalanced Text Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing multi-classification models, including prompt-based learning classifiers, struggle with imbalanced datasets, leading to performance drops due to unequal data distribution across classes, especially in text classification tasks.
Innovation Solution
A method involving a multi-stage prompt-based learning classifier that organizes datasets based on class frequency, performs reverse multi-stage training, and forward multi-stage inference, leveraging a pre-trained Language Model (LM) to handle imbalanced datasets efficiently and reduce model bias.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a pre-trained Language Model is used for text classification, then classification accuracy is improved, but computational cost increases
Solution Approach 1:
The patent segments the multi-class classification problem into multiple binary classification stages. Instead of using a single pre-trained LM for all classes, the system divides the task into sequential binary decisions, allowing lighter models to handle simpler discrimination tasks while maintaining overall accuracy.
Solution Approach 2:
The patent applies partial action by using simplified classification approaches for certain stages rather than applying the full pre-trained LM capacity to every classification decision. This reduces computational overhead while maintaining necessary accuracy through selective application of model complexity.
2Reliability
If ensemble machine learning techniques are used, then robustness in imbalanced class scenario is improved, but device complexity increases
Solution Approach 1:
The patent segments the complex ensemble problem into simpler sequential binary classification stages. Each stage handles a specific discrimination task, reducing the overall complexity while maintaining robustness through the staged approach to handling imbalanced classes.
3Adaptability or versatility
If prompt-based learning classifiers are used, then text classification capability is improved, but performance drops on imbalanced datasets
Solution Approach 1:
The patent segments the multi-class prompt-based classification into binary classification stages. This segmentation allows each stage to focus on distinguishing between two classes, improving performance on imbalanced datasets by avoiding the pitfalls of direct multi-class prompt-based approaches.
Solution Approach 2:
The patent inverts the traditional approach by using binary classification stages rather than direct multi-class prompt-based learning. This inversion allows the system to handle imbalanced data more effectively by breaking down the complex multi-class problem into manageable binary decisions.
Data Source
AI summary
One example method includes organizing, using prompt-classifiers (PC), an imbalanced dataset into ‘n’ different classes, and the organizing comprises performing a frequency analysis that identifies a respective number of samples in each of the ‘n’ different classes, and the organizing further comprises structuring, based on the frequency analysis, the imbalanced dataset using a cascaded one-versus-all approach to identify a target class and two remaining classes. Next, the method includes performing a reverse multi-stage prompt-classifier training process that comprises training the prompt-classifiers using the target classes and the two remaining classes, and the training is performed in reverse of an order in which the prompt-classifiers were used to organize the imbalanced dataset. Finally, the method includes performing an inferencing process using one or more of the prompt-classifiers, and the inferencing process continues until a then-current one of the prompt-classifiers correctly identifies the target class.