Unsupervised Non-transferable Text Classification

TL;DR

This paper introduces an unsupervised method for non-transferable text classification that prevents model transferability without requiring labeled target data, and includes a secret key mechanism for restoring access to the target domain.

Contribution

The paper proposes a novel unsupervised non-transferable learning approach with a secret key mechanism to recover target domain access, addressing limitations of prior supervised methods.

Findings

Effective in reducing transferability without labeled target data

Secret key mechanism successfully restores target domain access

Outperforms existing non-transferable learning methods

Abstract

Training a good deep learning model requires substantial data and computing resources, which makes the resulting neural model a valuable intellectual property. To prevent the neural network from being undesirably exploited, non-transferable learning has been proposed to reduce the model generalization ability in specific target domains. However, existing approaches require labeled data for the target domain which can be difficult to obtain. Furthermore, they do not have the mechanism to still recover the model's ability to access the target domain. In this paper, we propose a novel unsupervised non-transferable learning method for the text classification task that does not require annotated target domain data. We further introduce a secret key component in our approach for recovering the access to the target domain, where we design both an explicit and an implicit method for doing so. Extensive experiments demonstrate the effectiveness of our approach.