Figure 4. Overview of the software name disambiguation. For all pairs of extracted software entities (E₁, E₂), features are extracted (feature extraction) and used to determine a probability of linking (perceptron).

Finally, agglomerative clustering is performed to cluster similar software names.