Abstract:The goal of semi-supervised text classification (SSTC) is to train a model by exploring both a small number of labeled data and a large number of unlabeled data, such that the learned semi-supervised classifier performs better than the supervised classifier trained on solely the labeled samples. Pseudo-labeling is one of the most widely used SSTC techniques, which trains a teacher classifier with a small number of labeled examples to predict pseudo labels for the unlabeled data. The generated pseudo-labeled examples are then utilized to train a student classifier, such that the learned student classifier can outperform the teacher classifier. Nevertheless, the predicted pseudo labels may be inaccurate, making the performance of the student classifier degraded. The student classifier may perform even worse than the teacher classifier. To alleviate this issue, in this paper, we introduce a dual meta-learning ( DML ) technique for semi-supervised text classification, which improves the teacher and student classifiers simultaneously in an iterative manner. Specifically, we propose a meta-noise correction method to improve the student classifier by proposing a Noise Transition Matrix (NTM) with meta-learning to rectify the noisy pseudo labels. In addition, we devise a meta pseudo supervision method to improve the teacher classifier. Concretely, we exploit the feedback performance from the student classifier to further guide the teacher classifier to produce more accurate pseudo labels for the unlabeled data. In this way, both teacher and student classifiers can co-evolve in the iterative training process. Extensive experiments on four benchmark datasets highlight the effectiveness of our DML method against existing state-of-the-art methods for semi-supervised text classification. We release our code and data of this paper publicly at https://github.com/GRIT621/DML.

Improved multi-class semi-supervised learning with boosting method

Imbalanced Data Sets Classification Method Based on Over-Sampling Technique

Semi-supervised multi-class Adaboost by exploiting unlabeled data

Dual-Classifier Collaborative Method Based on Semi-Supervised Active Learning

Multi-class AdaBoost ELM

Optimally Combining Classifiers for Semi-Supervised Learning

sEMG multi-class pattern recognition based on semi-supervised boosting algorithm

Boosting Separability in Semisupervised Learning for Object Classification

Training Boosting-Like Algorithms with Semi-Supervised Subspace Learning

Chinese Short Text Categorization Based on Semi-Supervised Learning

Transfer Learning via Semi-supervised Boosting Method

Boost Picking: A Universal Method on Converting Supervised Classification to Semi-supervised Classification.

A Cluster-Based Semisupervised Ensemble for Multiclass Classification.

An Efficient Modified Boosting Method for Solving Classification Problems

Semi-Supervised Fuzzily Weighted Adaptive Boosting for Classification

Multi-Class L(2)-Boost With The Scoring Coding

Boosting Semi-Supervised Learning with Contrastive Complementary Labeling

Improved Boosting Algorithm for Small and Imbalanced Training Sets

Combined Classification Algorithm Based on Improved AdaBoost and SVM

AN ADAPTIVE BOOSTING ALGORITHM WITH SEMI-SUPERVISED LEARNING FOR FACIAL EXPRESSION RECOGNITION

Improving Semi-Supervised Text Classification with Dual Meta-Learning