Label Confusion Learning to Enhance Text Classification Models
This work addresses the problem of over-confident models and arbitrary predictions for text classification models, especially on confused or noisy datasets, by proposing a method to better represent label relationships.
The authors propose a Label Confusion Model (LCM) to address the limitations of one-hot label representations in text classification, particularly for confused or noisy datasets. LCM learns label confusion by calculating instance-label similarity during training, generating a more realistic label distribution to replace one-hot vectors, and improving classification performance on five benchmark datasets.
Representing a true label as a one-hot vector is a common practice in training text classification models. However, the one-hot representation may not adequately reflect the relation between the instances and labels, as labels are often not completely independent and instances may relate to multiple labels in practice. The inadequate one-hot representations tend to train the model to be over-confident, which may result in arbitrary prediction and model overfitting, especially for confused datasets (datasets with very similar labels) or noisy datasets (datasets with labeling errors). While training models with label smoothing (LS) can ease this problem in some degree, it still fails to capture the realistic relation among labels. In this paper, we propose a novel Label Confusion Model (LCM) as an enhancement component to current popular text classification models. LCM can learn label confusion to capture semantic overlap among labels by calculating the similarity between instances and labels during training and generate a better label distribution to replace the original one-hot label vector, thus improving the final classification performance. Extensive experiments on five text classification benchmark datasets reveal the effectiveness of LCM for several widely used deep learning classification models. Further experiments also verify that LCM is especially helpful for confused or noisy datasets and superior to the label smoothing method.