DisCo: Distilled Student Models Co-training for Semi-supervised Text Mining
This addresses the challenge of deploying efficient models in resource-constrained environments for text mining tasks, but it is incremental as it builds on existing distillation and co-training methods.
The paper tackles the problem of maintaining performance with lightweight models and limited labeled samples in text mining by introducing DisCo, a semi-supervised learning framework that uses knowledge distillation and co-training among student models, resulting in models that are 7.6 times smaller and 4.8 times faster while maintaining comparable performance.
Many text mining models are constructed by fine-tuning a large deep pre-trained language model (PLM) in downstream tasks. However, a significant challenge nowadays is maintaining performance when we use a lightweight model with limited labelled samples. We present DisCo, a semi-supervised learning (SSL) framework for fine-tuning a cohort of small student models generated from a large PLM using knowledge distillation. Our key insight is to share complementary knowledge among distilled student cohorts to promote their SSL effectiveness. DisCo employs a novel co-training technique to optimize a cohort of multiple small student models by promoting knowledge sharing among students under diversified views: model views produced by different distillation strategies and data views produced by various input augmentations. We evaluate DisCo on both semi-supervised text classification and extractive summarization tasks. Experimental results show that DisCo can produce student models that are 7.6 times smaller and 4.8 times faster in inference than the baseline PLMs while maintaining comparable performance. We also show that DisCo-generated student models outperform the similar-sized models elaborately tuned in distinct tasks.