Adaptive NER Data Augmentation for Minority Entity Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional named entity recognition (NER) models are biased towards majority class entities due to unbalanced representations in training datasets, leading to misclassification of minority class entities, and existing augmentation techniques further skew the datasets, impacting performance on minority classes.
Innovation Solution
Adaptive training data augmentation techniques calculate separate augmentation rates for each named entity category based on their distribution in the training data, generating adaptive numbers of additional utterances to normalize entity distributions and reduce bias, thereby improving NER model performance on minority classes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional NER models are trained on unbalanced training datasets, then the model can efficiently process majority class entities, but the model exhibits bias and misclassifies minority class entities
Solution Approach 1:
The patent applies local quality by implementing category-specific augmentation rates that treat different named entity categories differently. Instead of uniform data processing, the system calculates and applies customized augmentation rates for each entity category based on its distribution in the training data, thereby improving minority class representation while maintaining majority class performance
Solution Approach 2:
The patent changes the parameter of data augmentation by dynamically calculating augmentation rates based on entity category distribution. The system modifies the augmentation parameter (number of synthetic samples generated) according to each category's needs, using formulas that consider the proportion of each entity type in the training data to generate appropriate augmentation rates
2Quantity of substance
If existing data augmentation techniques are applied uniformly to all entity categories, then the overall training data volume increases, but the dataset becomes further skewed and performance on minority classes deteriorates
Solution Approach 1:
The patent implements local quality by differentiating augmentation strategies for each entity category. The system calculates separate augmentation rates for different categories based on their distribution characteristics, ensuring that minority classes receive appropriate augmentation while avoiding excessive augmentation that would skew the dataset further
Solution Approach 2:
The patent applies dynamics by making the augmentation rate adaptive and category-specific rather than static and uniform. The system dynamically calculates augmentation rates based on the distribution of each entity category in the training data, allowing the augmentation strategy to adapt to the specific needs of each category
Data Source
AI summary
Techniques are disclosed herein for adaptive training data augmentation to facilitate training named entity recognition (NER) models. Adaptive augmentation techniques are disclosed herein that take into consideration the distribution of different entity types within training data. The adaptive augmentation techniques generate adaptive numbers of augmented examples (e.g., utterances) based on the distribution of entities to make sure enough numbers of examples for minority class entities are generated during augmentation of the training data.