Fake scientific papers have recently become of interest within the academic community as a result of the identification of fake papers in the digital libraries of major academic publishers . Detecting and removing these papers is important for many reasons. We describe an investigation into the use of similarity search for detecting fake scientific papers by comparing several methods for signature construction and similarity scoring and describe a pseudo-relevance feedback technique that can be used to improve the effectiveness of these methods. Experiments on a dataset of 40,000 computer science papers show that precision, recall and MAP scores of 0.96, 0.99 and 0.99, respectively, can be achieved, thereby demonstrating the usefulness of similarity search in detecting fake scientific papers and ranking them highly.
展开▼
