An Evaluation of the GhostWriter System for Case-Based Content Suggestions

The Web has many sites where users can exchange goods and services. Often, the end-users must write free-text descriptions of the goods and services they have available, or the goods and services they are seeking. The quality of these descriptions is often low. In this paper, we describe the GhostWriter system, which encourages users to write descriptions that are more comprehensive. The system makes content suggestions from a case base of successful descriptions. The paper describes a new off-line ablation study that we have carried out to evaluate the system. The results show that GhostWriter has a high success rate in making suggestions that quickly recover ablated content.