Predicting Information Quality Flaws in Wikipedia by Using Classical and Deep Learning Approaches

Abstract

Quality flaws prediction in Wikipedia is an ongoing research trend. In particular, in this work we tackle the problem of automatically predicting five out of the ten most frequent quality flaws; namely: No footnotes, Notability, Primary Sources, Refimprove and Wikify. Different classical and deep learning state-of-the-art approaches were studied. From among the evaluated approaches, some of them always reach or improve the existing benchmarks on the test corpus from the $$1^{\mathrm {st}}$$ International Competition on Quality Flaw Prediction in Wikipedia; a well-known uniform evaluation corpus from this research field. Particularly, the results showed that under-bagged decision trees with different aggregation rules perform best improving the existing benchmarks for four out the five flaws.