Neural embedding-based machine learning models have shown promise for\npredicting novel links in knowledge graphs. Unfortunately, their practical\nutility is diminished by their lack of interpretability. Recently, the fully\ninterpretable, rule-based algorithm AnyBURL yielded highly competitive results\non many general-purpose link prediction benchmarks. However, current approaches\nfor aggregating predictions made by multiple rules are affected by\nredundancies. We improve upon AnyBURL by introducing the SAFRAN rule\napplication framework, which uses a novel aggregation approach called\nNon-redundant Noisy-OR that detects and clusters redundant rules prior to\naggregation. SAFRAN yields new state-of-the-art results for fully interpretable\nlink prediction on the established general-purpose benchmarks FB15K-237, WN18RR\nand YAGO3-10. Furthermore, it exceeds the results of multiple established\nembedding-based algorithms on FB15K-237 and WN18RR and narrows the gap between\nrule-based and embedding-based algorithms on YAGO3-10.\n