Gzip + KNN Outperforms BERT Classification Performance

Gzip + KNN Outperforms BERT Classification Performance

Paper Introduction “Low-Resource” Text Classification: A Parameter-Free Classification Method with Compressors https://aclanthology.org/2023.findings-acl.426/ This paper introduces a new method for text classification, providing a non-parametric alternative to Deep Neural Networks (DNNs). Although DNNs are widely used due to their high accuracy, they require a large amount of labeled data and millions of parameters, making their computational … Read more

Gzip + kNN Text Classification Beats Transformer with 14 Lines of Code

Gzip + kNN Text Classification Beats Transformer with 14 Lines of Code

A few days ago, the ACL 2023 awards were announced, attracting significant attention. Among the many papers included, one titled “Low-Resource Text Classification: A Parameter-Free Classification Method with Compressors” has started to generate much discussion. This paper was jointly completed by the University of Waterloo and AFAIK, but it is neither an award-winning paper nor … Read more