Assessing Document Sanitization for Controlled Information Release and Retrieval in Data Marketplaces

This study provides insights into both addressing data confidentiality concerns and enhancing document retrieval effectiveness in Data Marketplaces, which in this specific study consist of unstructured, textual documents. Through a semi-automatic sanitization process leveraging token masking with text summarization, possibly complemented by Coreference Resolution, the proposed solution mitigates the risk of inferring confidential information while maintaining search performance. Experimental results demonstrate encouraging improvements in both aspects with respect to baseline solutions.